diff options
| author | Linus Torvalds <torvalds@linux-foundation.org> | 2016-07-26 17:12:11 -0700 |
|---|---|---|
| committer | Linus Torvalds <torvalds@linux-foundation.org> | 2016-07-26 17:12:11 -0700 |
| commit | f7e68169941a26cb1ad764d53ef13721e6fe439a (patch) | |
| tree | cfcf45c75423c97c94c3932b2685890ded197711 | |
| parent | 3fc9d690936fb2e20e180710965ba2cc3a0881f8 (diff) | |
| parent | b5ab4a9ba5574430870391e93b125ada31217c4c (diff) | |
| download | linux-f7e68169941a26cb1ad764d53ef13721e6fe439a.tar.bz2 | |
Merge tag 'dm-4.8-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm
Pull device mapper updates from Mike Snitzer:
- initially based on Jens' 'for-4.8/core' (given all the flag churn)
and later merged with 'for-4.8/core' to pickup the QUEUE_FLAG_DAX
commits that DM depends on to provide its DAX support
- clean up the bio-based vs request-based DM core code by moving the
request-based DM core code out to dm-rq.[hc]
- reinstate bio-based support in the DM multipath target (done with the
idea that fast storage like NVMe over Fabrics could benefit) -- while
preserving support for request_fn and blk-mq request-based DM mpath
- SCSI and DM multipath persistent reservation fixes that were
coordinated with Martin Petersen.
- the DM raid target saw the most extensive change this cycle; it now
provides reshape and takeover support (by layering ontop of the
corresponding MD capabilities)
- DAX support for DM core and the linear, stripe and error targets
- a DM thin-provisioning block discard vs allocation race fix that
addresses potential for corruption
- a stable fix for DM verity-fec's block calculation during decode
- a few cleanups and fixes to DM core and various targets
* tag 'dm-4.8-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm: (73 commits)
dm: allow bio-based table to be upgraded to bio-based with DAX support
dm snap: add fake origin_direct_access
dm stripe: add DAX support
dm error: add DAX support
dm linear: add DAX support
dm: add infrastructure for DAX support
dm thin: fix a race condition between discarding and provisioning a block
dm btree: fix a bug in dm_btree_find_next_single()
dm raid: fix random optimal_io_size for raid0
dm raid: address checkpatch.pl complaints
dm: call PR reserve/unreserve on each underlying device
sd: don't use the ALL_TG_PT bit for reservations
dm: fix second blk_delay_queue() parameter to be in msec units not jiffies
dm raid: change logical functions to actually return bool
dm raid: use rdev_for_each in status
dm raid: use rs->raid_disks to avoid memory leaks on free
dm raid: support delta_disks for raid1, fix table output
dm raid: enhance reshape check and factor out reshape setup
dm raid: allow resize during recovery
dm raid: fix rs_is_recovering() to allow for lvextend
...
29 files changed, 4478 insertions, 1815 deletions
diff --git a/Documentation/device-mapper/dm-raid.txt b/Documentation/device-mapper/dm-raid.txt index df2d636b6088..e5b6497116f4 100644 --- a/Documentation/device-mapper/dm-raid.txt +++ b/Documentation/device-mapper/dm-raid.txt @@ -14,8 +14,12 @@ The target is named "raid" and it accepts the following parameters: <#raid_devs> <metadata_dev0> <dev0> [.. <metadata_devN> <devN>] <raid_type>: + raid0 RAID0 striping (no resilience) raid1 RAID1 mirroring - raid4 RAID4 dedicated parity disk + raid4 RAID4 with dedicated last parity disk + raid5_n RAID5 with dedicated last parity disk suporting takeover + Same as raid4 + -Transitory layout raid5_la RAID5 left asymmetric - rotating parity 0 with data continuation raid5_ra RAID5 right asymmetric @@ -30,7 +34,19 @@ The target is named "raid" and it accepts the following parameters: - rotating parity N (right-to-left) with data restart raid6_nc RAID6 N continue - rotating parity N (right-to-left) with data continuation + raid6_n_6 RAID6 with dedicate parity disks + - parity and Q-syndrome on the last 2 disks; + laylout for takeover from/to raid4/raid5_n + raid6_la_6 Same as "raid_la" plus dedicated last Q-syndrome disk + - layout for takeover from raid5_la from/to raid6 + raid6_ra_6 Same as "raid5_ra" dedicated last Q-syndrome disk + - layout for takeover from raid5_ra from/to raid6 + raid6_ls_6 Same as "raid5_ls" dedicated last Q-syndrome disk + - layout for takeover from raid5_ls from/to raid6 + raid6_rs_6 Same as "raid5_rs" dedicated last Q-syndrome disk + - layout for takeover from raid5_rs from/to raid6 raid10 Various RAID10 inspired algorithms chosen by additional params + (see raid10_format and raid10_copies below) - RAID10: Striped Mirrors (aka 'Striping on top of mirrors') - RAID1E: Integrated Adjacent Stripe Mirroring - RAID1E: Integrated Offset Stripe Mirroring @@ -116,10 +132,41 @@ The target is named "raid" and it accepts the following parameters: Here we see layouts closely akin to 'RAID1E - Integrated Offset Stripe Mirroring'. + [delta_disks <N>] + The delta_disks option value (-251 < N < +251) triggers + device removal (negative value) or device addition (positive + value) to any reshape supporting raid levels 4/5/6 and 10. + RAID levels 4/5/6 allow for addition of devices (metadata + and data device tupel), raid10_near and raid10_offset only + allow for device addtion. raid10_far does not support any + reshaping at all. + A minimum of devices have to be kept to enforce resilience, + which is 3 devices for raid4/5 and 4 devices for raid6. + + [data_offset <sectors>] + This option value defines the offset into each data device + where the data starts. This is used to provide out-of-place + reshaping space to avoid writing over data whilst + changing the layout of stripes, hence an interruption/crash + may happen at any time without the risk of losing data. + E.g. when adding devices to an existing raid set during + forward reshaping, the out-of-place space will be allocated + at the beginning of each raid device. The kernel raid4/5/6/10 + MD personalities supporting such device addition will read the data from + the existing first stripes (those with smaller number of stripes) + starting at data_offset to fill up a new stripe with the larger + number of stripes, calculate the redundancy blocks (CRC/Q-syndrome) + and write that new stripe to offset 0. Same will be applied to all + N-1 other new stripes. This out-of-place scheme is used to change + the RAID type (i.e. the allocation algorithm) as well, e.g. + changing from raid5_ls to raid5_n. + <#raid_devs>: The number of devices composing the array. Each device consists of two entries. The first is the device containing the metadata (if any); the second is the one containing the - data. + data. A Maximum of 64 metadata/data device entries are supported + up to target version 1.8.0. + 1.9.0 supports up to 253 which is enforced by the used MD kernel runtime. If a drive has failed or is missing at creation time, a '-' can be given for both the metadata and data drives for a given position. @@ -207,7 +254,6 @@ include: "recover"- Initiate/continue a recover process. "check" - Initiate a check (i.e. a "scrub") of the array. "repair" - Initiate a repair of the array. - "reshape"- Currently unsupported (-EINVAL). Discard Support @@ -257,3 +303,9 @@ Version History 1.5.2 'mismatch_cnt' is zero unless [last_]sync_action is "check". 1.6.0 Add discard support (and devices_handle_discard_safely module param). 1.7.0 Add support for MD RAID0 mappings. +1.8.0 Explictely check for compatible flags in the superblock metadata + and reject to start the raid set if any are set by a newer + target version, thus avoiding data corruption on a raid set + with a reshape in progress. +1.9.0 Add support for RAID level takeover/reshape/region size + and set size reduction. diff --git a/drivers/md/Makefile b/drivers/md/Makefile index 52ba8dd82821..3cbda1af87a0 100644 --- a/drivers/md/Makefile +++ b/drivers/md/Makefile @@ -3,7 +3,8 @@ # dm-mod-y += dm.o dm-table.o dm-target.o dm-linear.o dm-stripe.o \ - dm-ioctl.o dm-io.o dm-kcopyd.o dm-sysfs.o dm-stats.o + dm-ioctl.o dm-io.o dm-kcopyd.o dm-sysfs.o dm-stats.o \ + dm-rq.o dm-multipath-y += dm-path-selector.o dm-mpath.o dm-snapshot-y += dm-snap.o dm-exception-store.o dm-snap-transient.o \ dm-snap-persistent.o diff --git a/drivers/md/dm-builtin.c b/drivers/md/dm-builtin.c index 6c9049c51b2b..f092771878c2 100644 --- a/drivers/md/dm-builtin.c +++ b/drivers/md/dm-builtin.c @@ -1,4 +1,4 @@ -#include "dm.h" +#include "dm-core.h" /* * The kobject release method must not be placed in the module itself, diff --git a/drivers/md/dm-core.h b/drivers/md/dm-core.h new file mode 100644 index 000000000000..40ceba1fe8be --- /dev/null +++ b/drivers/md/dm-core.h @@ -0,0 +1,149 @@ +/* + * Internal header file _only_ for device mapper core + * + * Copyright (C) 2016 Red Hat, Inc. All rights reserved. + * + * This file is released under the LGPL. + */ + +#ifndef DM_CORE_INTERNAL_H +#define DM_CORE_INTERNAL_H + +#include <linux/kthread.h> +#include <linux/ktime.h> +#include <linux/blk-mq.h> + +#include <trace/events/block.h> + +#include "dm.h" + +#define DM_RESERVED_MAX_IOS 1024 + +struct dm_kobject_holder { + struct kobject kobj; + struct completion completion; +}; + +/* + * DM core internal structure that used directly by dm.c and dm-rq.c + * DM targets must _not_ deference a mapped_device to directly access its members! + */ +struct mapped_device { + struct srcu_struct io_barrier; + struct mutex suspend_lock; + + /* + * The current mapping (struct dm_table *). + * Use dm_get_live_table{_fast} or take suspend_lock for + * dereference. + */ + void __rcu *map; + + struct list_head table_devices; + struct mutex table_devices_lock; + + unsigned long flags; + + struct request_queue *queue; + int numa_node_id; + + unsigned type; + /* Protect queue and type against concurrent access. */ + struct mutex type_lock; + + atomic_t holders; + atomic_t open_count; + + struct dm_target *immutable_target; + struct target_type *immutable_target_type; + + struct gendisk *disk; + char name[16]; + + void *interface_ptr; + + /* + * A list of ios that arrived while we were suspended. + */ + atomic_t pending[2]; + wait_queue_head_t wait; + struct work_struct work; + spinlock_t deferred_lock; + struct bio_list deferred; + + /* + * Event handling. + */ + wait_queue_head_t eventq; + atomic_t event_nr; + atomic_t uevent_seq; + struct list_head uevent_list; + spinlock_t uevent_lock; /* Protect access to uevent_list */ + + /* the number of internal suspends */ + unsigned internal_suspend_count; + + /* + * Processing queue (flush) + */ + struct workqueue_struct *wq; + + /* + * io objects are allocated from here. + */ + mempool_t *io_pool; + mempool_t *rq_pool; + + struct bio_set *bs; + + /* + * freeze/thaw support require holding onto a super block + */ + struct super_block *frozen_sb; + + /* forced geometry settings */ + struct hd_geometry geometry; + + struct block_device *bdev; + + /* kobject and completion */ + struct dm_kobject_holder kobj_holder; + + /* zero-length flush that will be cloned and submitted to targets */ + struct bio flush_bio; + + struct dm_stats stats; + + struct kthread_worker kworker; + struct task_struct *kworker_task; + + /* for request-based merge heuristic in dm_request_fn() */ + unsigned seq_rq_merge_deadline_usecs; + int last_rq_rw; + sector_t last_rq_pos; + ktime_t last_rq_start_time; + + /* for blk-mq request-based DM support */ + struct blk_mq_tag_set *tag_set; + bool use_blk_mq:1; + bool init_tio_pdu:1; +}; + +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); + +static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj) +{ + return &container_of(kobj, struct dm_kobject_holder, kobj)->completion; +} + +unsigned __dm_get_module_param(unsigned *module_param, unsigned def, unsigned max); + +static inline bool dm_message_test_buffer_overflow(char *result, unsigned maxlen) +{ + return !maxlen || strlen(result) + 1 >= maxlen; +} + +#endif diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c index 96dd5d7e454a..8f2e3e2ffd26 100644 --- a/drivers/md/dm-crypt.c +++ b/drivers/md/dm-crypt.c @@ -683,7 +683,7 @@ static int crypt_iv_tcw_whitening(struct crypt_config *cc, u8 *data) { struct iv_tcw_private *tcw = &cc->iv_gen_private.tcw; - u64 sector = cpu_to_le64((u64)dmreq->iv_sector); + __le64 sector = cpu_to_le64(dmreq->iv_sector); u8 buf[TCW_WHITENING_SIZE]; SHASH_DESC_ON_STACK(desc, tcw->crc32_tfm); int i, r; @@ -722,7 +722,7 @@ static int crypt_iv_tcw_gen(struct crypt_config *cc, u8 *iv, struct dm_crypt_request *dmreq) { struct iv_tcw_private *tcw = &cc->iv_gen_private.tcw; - u64 sector = cpu_to_le64((u64)dmreq->iv_sector); + __le64 sector = cpu_to_le64(dmreq->iv_sector); u8 *src; int r = 0; diff --git a/drivers/md/dm-io.c b/drivers/md/dm-io.c index 0e225fd4a8d1..daa03e41654a 100644 --- a/drivers/md/dm-io.c +++ b/drivers/md/dm-io.c @@ -5,7 +5,7 @@ * This file is released under the GPL. */ -#include "dm.h" +#include "dm-core.h" #include <linux/device-mapper.h> diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c index 2c7ca258c4e4..966eb4b61aed 100644 --- a/drivers/md/dm-ioctl.c +++ b/drivers/md/dm-ioctl.c @@ -5,7 +5,7 @@ * This file is released under the GPL. */ -#include "dm.h" +#include "dm-core.h" #include <linux/module.h> #include <linux/vmalloc.h> @@ -1267,6 +1267,15 @@ static int populate_table(struct dm_table *table, return dm_table_complete(table); } +static bool is_valid_type(unsigned cur, unsigned new) +{ + if (cur == new || + (cur == DM_TYPE_BIO_BASED && new == DM_TYPE_DAX_BIO_BASED)) + return true; + + return false; +} + static int table_load(struct dm_ioctl *param, size_t param_size) { int r; @@ -1309,7 +1318,7 @@ static int table_load(struct dm_ioctl *param, size_t param_size) DMWARN("unable to set up device queue for new table."); goto err_unlock_md_type; } - } else if (dm_get_md_type(md) != dm_table_get_type(t)) { + } else if (!is_valid_type(dm_get_md_type(md), dm_table_get_type(t))) { DMWARN("can't change device type after initial table load."); r = -EINVAL; goto err_unlock_md_type; @@ -1670,8 +1679,7 @@ static int check_version(unsigned int cmd, struct dm_ioctl __user *user) return r; } -#define DM_PARAMS_KMALLOC 0x0001 /* Params alloced with kmalloc */ -#define DM_PARAMS_VMALLOC 0x0002 /* Params alloced with vmalloc */ +#define DM_PARAMS_MALLOC 0x0001 /* Params allocated with kvmalloc() */ #define DM_WIPE_BUFFER 0x0010 /* Wipe input buffer before returning from ioctl */ static void free_params(struct dm_ioctl *param, size_t param_size, int param_flags) @@ -1679,10 +1687,8 @@ static void free_params(struct dm_ioctl *param, size_t param_size, int param_fla if (param_flags & DM_WIPE_BUFFER) memset(param, 0, param_size); - if (param_flags & DM_PARAMS_KMALLOC) - kfree(param); - if (param_flags & DM_PARAMS_VMALLOC) - vfree(param); + if (param_flags & DM_PARAMS_MALLOC) + kvfree(param); } static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kernel, @@ -1714,19 +1720,14 @@ static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kern * Use kmalloc() rather than vmalloc() when we can. */ dmi = NULL; - if (param_kernel->data_size <= KMALLOC_MAX_SIZE) { + if (param_kernel->data_size <= KMALLOC_MAX_SIZE) dmi = kmalloc(param_kernel->data_size, GFP_NOIO | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN); - if (dmi) - *param_flags |= DM_PARAMS_KMALLOC; - } if (!dmi) { unsigned noio_flag; noio_flag = memalloc_noio_save(); dmi = __vmalloc(param_kernel->data_size, GFP_NOIO | __GFP_HIGH | __GFP_HIGHMEM, PAGE_KERNEL); memalloc_noio_restore(noio_flag); - if (dmi) - *param_flags |= DM_PARAMS_VMALLOC; } if (!dmi) { @@ -1735,6 +1736,8 @@ static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kern return -ENOMEM; } + *param_flags |= DM_PARAMS_MALLOC; + if (copy_from_user(dmi, user, param_kernel->data_size)) goto bad; diff --git a/drivers/md/dm-kcopyd.c b/drivers/md/dm-kcopyd.c index 9da1d54ac6cb..9e9d04cb7d51 100644 --- a/drivers/md/dm-kcopyd.c +++ b/drivers/md/dm-kcopyd.c @@ -26,7 +26,7 @@ #include <linux/device-mapper.h> #include <linux/dm-kcopyd.h> -#include "dm.h" +#include "dm-core.h" #define SUB_JOB_SIZE 128 #define SPLIT_COUNT 8 diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c index 05c35aacb3aa..6d35dd4e9efb 100644 --- a/drivers/md/dm-linear.c +++ b/drivers/md/dm-linear.c @@ -141,9 +141,27 @@ static int linear_iterate_devices(struct dm_target *ti, return fn(ti, lc->dev, lc->start, ti->len, data); } +static long linear_direct_access(struct dm_target *ti, sector_t sector, + void __pmem **kaddr, pfn_t *pfn, long size) +{ + struct linear_c *lc = ti->private; + struct block_device *bdev = lc->dev->bdev; + struct blk_dax_ctl dax = { + .sector = linear_map_sector(ti, sector), + .size = size, + }; + long ret; + + ret = bdev_direct_access(bdev, &dax); + *kaddr = dax.addr; + *pfn = dax.pfn; + + return ret; +} + static struct target_type linear_target = { .name = "linear", - .version = {1, 2, 1}, + .version = {1, 3, 0}, .module = THIS_MODULE, .ctr = linear_ctr, .dtr = linear_dtr, @@ -151,6 +169,7 @@ static struct target_type linear_target = { .status = linear_status, .prepare_ioctl = linear_prepare_ioctl, .iterate_devices = linear_iterate_devices, + .direct_access = linear_direct_access, }; int __init dm_linear_init(void) diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c index 52baf8a5b0f4..7eac080fcb18 100644 --- a/drivers/md/dm-mpath.c +++ b/drivers/md/dm-mpath.c @@ -7,7 +7,8 @@ #include <linux/device-mapper.h> -#include "dm.h" +#include "dm-rq.h" +#include "dm-bio-record.h" #include "dm-path-selector.h" #include "dm-uevent.h" @@ -89,6 +90,8 @@ struct multipath { atomic_t pg_init_in_progress; /* Only one pg_init allowed at once */ atomic_t pg_init_count; /* Number of times pg_init called */ + unsigned queue_mode; + /* * We must use a mempool of dm_mpath_io structs so that we * can resubmit bios on error. @@ -97,10 +100,13 @@ struct multipath { struct mutex work_mutex; struct work_struct trigger_event; + + struct work_struct process_queued_bios; + struct bio_list queued_bios; }; /* - * Context information attached to each bio we process. + * Context information attached to each io we process. */ struct dm_mpath_io { struct pgpath *pgpath; @@ -114,6 +120,7 @@ static struct kmem_cache *_mpio_cache; static struct workqueue_struct *kmultipathd, *kmpath_handlerd; static void trigger_event(struct work_struct *work); static void activate_path(struct work_struct *work); +static void process_queued_bios(struct work_struct *work); /*----------------------------------------------- * Multipath state flags. @@ -185,7 +192,7 @@ static void free_priority_group(struct priority_group *pg, kfree(pg); } -static struct multipath *alloc_multipath(struct dm_target *ti, bool use_blk_mq) +static struct multipath *alloc_multipath(struct dm_target *ti) { struct multipath *m; @@ -203,15 +210,7 @@ static struct multipath *alloc_multipath(struct dm_target *ti, bool use_blk_mq) mutex_init(&m->work_mutex); m->mpio_pool = NULL; - if (!use_blk_mq) { - unsigned min_ios = dm_get_reserved_rq_based_ios(); - - m->mpio_pool = mempool_create_slab_pool(min_ios, _mpio_cache); - if (!m->mpio_pool) { - kfree(m); - return NULL; - } - } + m->queue_mode = DM_TYPE_NONE; m->ti = ti; ti->private = m; @@ -220,6 +219,39 @@ static struct multipath *alloc_multipath(struct dm_target *ti, bool use_blk_mq) return m; } +static int alloc_multipath_stage2(struct dm_target *ti, struct multipath *m) +{ + if (m->queue_mode == DM_TYPE_NONE) { + /* + * Default to request-based. + */ + if (dm_use_blk_mq(dm_table_get_md(ti->table))) + m->queue_mode = DM_TYPE_MQ_REQUEST_BASED; + else + m->queue_mode = DM_TYPE_REQUEST_BASED; + } + + if (m->queue_mode == DM_TYPE_REQUEST_BASED) { + unsigned min_ios = dm_get_reserved_rq_based_ios(); + + m->mpio_pool = mempool_create_slab_pool(min_ios, _mpio_cache); + if (!m->mpio_pool) + return -ENOMEM; + } + else if (m->queue_mode == DM_TYPE_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); + } + + dm_table_set_type(ti->table, m->queue_mode); + + return 0; +} + static void free_multipath(struct multipath *m) { struct priority_group *pg, *tmp; @@ -272,6 +304,41 @@ static void clear_request_fn_mpio(struct multipath *m, union map_info *info) } } +static size_t multipath_per_bio_data_size(void) +{ + return sizeof(struct dm_mpath_io) + sizeof(struct dm_bio_details); +} + +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) +{ + /* 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) +{ + struct dm_mpath_io *mpio = get_mpio_from_bio(bio); + struct dm_bio_details *bio_details = get_bio_details_from_bio(bio); + + memset(mpio, 0, sizeof(*mpio)); + memset(bio_details, 0, sizeof(*bio_details)); + dm_bio_record(bio_details, bio); + + if (mpio_p) + *mpio_p = mpio; + if (bio_details_p) + *bio_details_p = bio_details; +} + /*----------------------------------------------- * Path selection *-----------------------------------------------*/ @@ -431,16 +498,26 @@ failed: * and multipath_resume() calls and we have no need to check * for the DMF_NOFLUSH_SUSPENDING flag. */ -static int must_push_back(struct multipath *m) +static bool __must_push_back(struct multipath *m) +{ + return ((test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) != + test_bit(MPATHF_SAVED_QUEUE_IF_NO_PATH, &m->flags)) && + dm_noflush_suspending(m->ti)); +} + +static bool must_push_back_rq(struct multipath *m) { return (test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) || - ((test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) != - test_bit(MPATHF_SAVED_QUEUE_IF_NO_PATH, &m->flags)) && - dm_noflush_suspending(m->ti))); + __must_push_back(m)); +} + +static bool must_push_back_bio(struct multipath *m) +{ + return __must_push_back(m); } /* - * Map cloned requests + * Map cloned requests (request-based multipath) */ static int __multipath_map(struct dm_target *ti, struct request *clone, union map_info *map_context, @@ -459,7 +536,7 @@ static int __multipath_map(struct dm_target *ti, struct request *clone, pgpath = choose_pgpath(m, nr_bytes); if (!pgpath) { - if (!must_push_back(m)) + if (!must_push_back_rq(m)) r = -EIO; /* Failed */ return r; } else if (test_bit(MPATHF_QUEUE_IO, &m->flags) || @@ -530,6 +607,108 @@ 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) +{ + size_t nr_bytes = bio->bi_iter.bi_size; + struct pgpath *pgpath; + unsigned long flags; + bool queue_io; + + /* Do we need to select a new pgpath? */ + pgpath = lockless_dereference(m->current_pgpath); + queue_io = test_bit(MPATHF_QUEUE_IO, &m->flags); + if (!pgpath || !queue_io) + pgpath = choose_pgpath(m, nr_bytes); + + if ((pgpath && queue_io) || + (!pgpath && 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); + /* 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; + } + + if (!pgpath) { + if (!must_push_back_bio(m)) + return -EIO; + return DM_MAPIO_REQUEUE; + } + + mpio->pgpath = pgpath; + mpio->nr_bytes = nr_bytes; + + bio->bi_error = 0; + bio->bi_bdev = pgpath->path.dev->bdev; + bio->bi_rw |= REQ_FAILFAST_TRANSPORT; + + if (pgpath->pg->ps.type->start_io) + pgpath->pg->ps.type->start_io(&pgpath->pg->ps, + &pgpath->path, + nr_bytes); + return DM_MAPIO_REMAPPED; +} + +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); + + return __multipath_map_bio(m, bio, mpio); +} + +static void process_queued_bios_list(struct multipath *m) +{ + if (m->queue_mode == DM_TYPE_BIO_BASED) + queue_work(kmultipathd, &m->process_queued_bios); +} + +static void process_queued_bios(struct work_struct *work) +{ + int r; + unsigned long flags; + struct bio *bio; + struct bio_list bios; + struct blk_plug plug; + struct multipath *m = + container_of(work, struct multipath, process_queued_bios); + + bio_list_init(&bios); + + spin_lock_irqsave(&m->lock, flags); + + if (bio_list_empty(&m->queued_bios)) { + spin_unlock_irqrestore(&m->lock, flags); + return; + } + + bio_list_merge(&bios, &m->queued_bios); + bio_list_init(&m->queued_bios); + + spin_unlock_irqrestore(&m->lock, flags); + + blk_start_plug(&plug); + while ((bio = bio_list_pop(&bios))) { + r = __multipath_map_bio(m, bio, get_mpio_from_bio(bio)); + if (r < 0 || r == DM_MAPIO_REQUEUE) { + bio->bi_error = r; + bio_endio(bio); + } else if (r == DM_MAPIO_REMAPPED) + generic_make_request(bio); + } + blk_finish_plug(&plug); +} + +/* * If we run out of usable paths, should we queue I/O or error it? */ static int queue_if_no_path(struct multipath *m, bool queue_if_no_path, @@ -557,8 +736,10 @@ static int queue_if_no_path(struct multipath *m, bool queue_if_no_path, spin_unlock_irqrestore(&m->lock, flags); - if (!queue_if_no_path) + if (!queue_if_no_path) { dm_table_run_md_queue_async(m->ti->table); + process_queued_bios_list(m); + } return 0; } @@ -798,6 +979,12 @@ 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) { + dm_consume_args(as, hw_argc); + DMERR("bio-based multipath doesn't allow hardware handler args"); + return 0; + } + m->hw_handler_name = kstrdup(dm_shift_arg(as), GFP_KERNEL); if (hw_argc > 1) { @@ -833,7 +1020,7 @@ static int parse_features(struct dm_arg_set *as, struct multipath *m) const char *arg_name; static struct dm_arg _args[] = { - {0, 6, "invalid number of feature args"}, + {0, 8, "invalid number of feature args"}, {1, 50, "pg_init_retries must be between 1 and 50"}, {0, 60000, "pg_init_delay_msecs must be between 0 and 60000"}, }; @@ -873,6 +1060,24 @@ static int parse_features(struct dm_arg_set *as, struct multipath *m) continue; } + if (!strcasecmp(arg_name, "queue_mode") && + (argc >= 1)) { + const char *queue_mode_name = dm_shift_arg(as); + + if (!strcasecmp(queue_mode_name, "bio")) + m->queue_mode = DM_TYPE_BIO_BASED; + else if (!strcasecmp(queue_mode_name, "rq")) + m->queue_mode = DM_TYPE_REQUEST_BASED; + else if (!strcasecmp(queue_mode_name, "mq")) + m->queue_mode = DM_TYPE_MQ_REQUEST_BASED; + else { + ti->error = "Unknown 'queue_mode' requested"; + r = -EINVAL; + } + argc--; + continue; + } + ti->error = "Unrecognised multipath feature request"; r = -EINVAL; } while (argc && !r); @@ -880,8 +1085,7 @@ static int parse_features(struct dm_arg_set *as, struct multipath *m) return r; } -static int multipath_ctr(struct dm_target *ti, unsigned int argc, - char **argv) +static int multipath_ctr(struct dm_target *ti, unsigned argc, char **argv) { /* target arguments */ static struct dm_arg _args[] = { @@ -894,12 +1098,11 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc, struct dm_arg_set as; unsigned pg_count = 0; unsigned next_pg_num; - bool use_blk_mq = dm_use_blk_mq(dm_table_get_md(ti->table)); as.argc = argc; as.argv = argv; - m = alloc_multipath(ti, use_blk_mq); + m = alloc_multipath(ti); if (!m) { ti->error = "can't allocate multipath"; return -EINVAL; @@ -909,6 +1112,10 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc, if (r) goto bad; + r = alloc_multipath_stage2(ti, m); + if (r) + goto bad; + r = parse_hw_handler(&as, m); if (r) goto bad; @@ -958,7 +1165,9 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc, ti->num_flush_bios = 1; ti->num_discard_bios = 1; ti->num_write_same_bios = 1; - if (use_blk_mq) + if (m->queue_mode == DM_TYPE_BIO_BASED) + ti->per_io_data_size = multipath_per_bio_data_size(); + else if (m->queue_mode == DM_TYPE_MQ_REQUEST_BASED) ti->per_io_data_size = sizeof(struct dm_mpath_io); return 0; @@ -1083,8 +1292,10 @@ static int reinstate_path(struct pgpath *pgpath) out: spin_unlock_irqrestore(&m->lock, flags); - if (run_queue) + if (run_queue) { dm_table_run_md_queue_async(m->ti->table); + process_queued_bios_list(m); + } return r; } @@ -1281,6 +1492,8 @@ static void pg_init_done(void *data, int errors) } clear_bit(MPATHF_QUEUE_IO, &m->flags); + process_queued_bios_list(m); + /* * Wake up any thread waiting to suspend. */ @@ -1328,7 +1541,7 @@ static int do_end_io(struct multipath *m, struct request *clone, * during end I/O handling, since those clone requests don't have * bio clones. If we queue them inside the multipath target, * we need to make bio clones, that requires memory allocation. - * (See drivers/md/dm.c:end_clone_bio() about why the clone requests + * (See drivers/md/dm-rq.c:end_clone_bio() about why the clone requests * don't have bio clones.) * Instead of queueing the clone request here, we queue the original * request into dm core, which will remake a clone request and @@ -1347,7 +1560,7 @@ static int do_end_io(struct multipath *m, struct request *clone, if (!atomic_read(&m->nr_valid_paths)) { if (!test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags)) { - if (!must_push_back(m)) + if (!must_push_back_rq(m)) r = -EIO; } else { if (error == -EBADE) @@ -1381,6 +1594,64 @@ static int multipath_end_io(struct dm_target *ti, struct request *clone, return r; } +static int do_end_io_bio(struct multipath *m, struct bio *clone, + int error, struct dm_mpath_io *mpio) +{ + unsigned long flags; + + if (!error) + return 0; /* I/O complete */ + + if (noretry_error(error)) + return error; + + if (mpio->pgpath) + fail_path(mpio->pgpath); + + if (!atomic_read(&m->nr_valid_paths)) { + if (!test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags)) { + if (!must_push_back_bio(m)) + return -EIO; + return DM_ENDIO_REQUEUE; + } else { + if (error == -EBADE) + return error; + } + } + + /* 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); + if (!test_bit(MPATHF_QUEUE_IO, &m->flags)) + queue_work(kmultipathd, &m->process_queued_bios); + + return DM_ENDIO_INCOMPLETE; +} + +static int multipath_end_io_bio(struct dm_target *ti, struct bio *clone, int error) +{ + struct multipath *m = ti->private; + struct dm_mpath_io *mpio = get_mpio_from_bio(clone); + struct pgpath *pgpath; + struct path_selector *ps; + int r; + + BUG_ON(!mpio); + + r = do_end_io_bio(m, clone, error, mpio); + pgpath = mpio->pgpath; + if (pgpath) { + ps = &pgpath->pg->ps; + if (ps->type->end_io) + ps->type->end_io(ps, &pgpath->path, mpio->nr_bytes); + } + + return r; +} + /* * Suspend can't complete until all the I/O is processed so if * the last path fails we must error any remaining I/O. @@ -1454,7 +1725,9 @@ static void multipath_status(struct dm_target *ti, status_type_t type, DMEMIT("%u ", test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags) + (m->pg_init_retries > 0) * 2 + (m->pg_init_delay_msecs != DM_PG_INIT_DELAY_DEFAULT) * 2 + - test_bit(MPATHF_RETAIN_ATTACHED_HW_HANDLER, &m->flags)); + test_bit(MPATHF_RETAIN_ATTACHED_HW_HANDLER, &m->flags) + + (m->queue_mode != DM_TYPE_REQUEST_BASED) * 2); + if (test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags)) DMEMIT("queue_if_no_path "); if (m->pg_init_retries) @@ -1463,6 +1736,16 @@ static void multipath_status(struct dm_target *ti, status_type_t type, DMEMIT("pg_init_delay_msecs %u ", m->pg_init_delay_msecs); if (test_bit(MPATHF_RETAIN_ATTACHED_HW_HANDLER, &m->flags)) DMEMIT("retain_attached_hw_handler "); + if (m->queue_mode != DM_TYPE_REQUEST_BASED) { + switch(m->queue_mode) { + case DM_TYPE_BIO_BASED: + DMEMIT("queue_mode bio "); + break; + case DM_TYPE_MQ_REQUEST_BASED: + DMEMIT("queue_mode mq "); + break; + } + } } if (!m->hw_handler_name || type == STATUSTYPE_INFO) @@ -1642,6 +1925,7 @@ static int multipath_prepare_ioctl(struct dm_target *ti, if (test_bit(MPATHF_PG_INIT_REQUIRED, &m->flags)) pg_init_all_paths(m); dm_table_run_md_queue_async(m->ti->table); + process_queued_bios_list(m); } /* @@ -1748,7 +2032,7 @@ static int multipath_busy(struct dm_target *ti) *---------------------------------------------------------------*/ static struct target_type multipath_target = { .name = "multipath", - .version = {1, 11, 0}, + .version = {1, 12, 0}, .features = DM_TARGET_SINGLETON | DM_TARGET_IMMUTABLE, .module = THIS_MODULE, .ctr = multipath_ctr, @@ -1757,6 +2041,8 @@ static struct target_type multipath_target = { .clone_and_map_rq = multipath_clone_and_map, .release_clone_rq = multipath_release_clone, .rq_end_io = multipath_end_io, + .map = multipath_map_bio, + .end_io = multipath_end_io_bio, .presuspend = multipath_presuspend, .postsuspend = multipath_postsuspend, .resume = multipath_resume, @@ -1771,14 +2057,14 @@ static int __init dm_multipath_init(void) { int r; - /* allocate a slab for the dm_ios */ + /* allocate a slab for the dm_mpath_ios */ _mpio_cache = KMEM_CACHE(dm_mpath_io, 0); if (!_mpio_cache) return -ENOMEM; r = dm_register_target(&multipath_target); if (r < 0) { - DMERR("register failed %d", r); + DMERR("request-based register failed %d", r); r = -EINVAL; goto bad_register_target; } @@ -1804,10 +2090,6 @@ static int __init dm_multipath_init(void) goto bad_alloc_kmpath_handlerd; } - DMINFO("version %u.%u.%u loaded", - multipath_target.version[0], multipath_target.version[1], - multipath_target.version[2]); - return 0; bad_alloc_kmpath_handlerd: diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c index 8cbac62b1602..84983549b5e1 100644 --- a/drivers/md/dm-raid.c +++ b/drivers/md/dm-raid.c @@ -1,6 +1,6 @@ /* * Copyright (C) 2010-2011 Neil Brown - * Copyright (C) 2010-2015 Red Hat, Inc. All rights reserved. + * Copyright (C) 2010-2016 Red Hat, Inc. All rights reserved. * * This file is released under the GPL. */ @@ -17,7 +17,12 @@ #include <linux/device-mapper.h> #define DM_MSG_PREFIX "raid" -#define MAX_RAID_DEVICES 253 /* raid4/5/6 limit */ +#define MAX_RAID_DEVICES 253 /* md-raid kernel limit */ + +/* + * Minimum sectors of free reshape space per raid device + */ +#define MIN_FREE_RESHAPE_SPACE to_sector(4*4096) static bool devices_handle_discard_safely = false; @@ -25,12 +30,12 @@ static bool devices_handle_discard_safely = false; * The following flags are used by dm-raid.c to set up the array state. * They must be cleared before md_run is called. */ -#define FirstUse 10 /* rdev flag */ +#define FirstUse 10 /* rdev flag */ struct raid_dev { /* * Two DM devices, one to hold metadata and one to hold the - * actual data/parity. The reason for this is to not confuse + * actual data/parity. The reason for this is to not confuse * ti->len and give more flexibility in altering size and * characteristics. * @@ -46,25 +51,175 @@ struct raid_dev { }; /* + * Bits for establishing rs->ctr_flags + * + * 1 = no flag value + * 2 = flag with value + */ +#define __CTR_FLAG_SYNC 0 /* 1 */ /* Not with raid0! */ +#define __CTR_FLAG_NOSYNC 1 /* 1 */ /* Not with raid0! */ +#define __CTR_FLAG_REBUILD 2 /* 2 */ /* Not with raid0! */ +#define __CTR_FLAG_DAEMON_SLEEP 3 /* 2 */ /* Not with raid0! */ +#define __CTR_FLAG_MIN_RECOVERY_RATE 4 /* 2 */ /* Not with raid0! */ +#define __CTR_FLAG_MAX_RECOVERY_RATE 5 /* 2 */ /* Not with raid0! */ +#define __CTR_FLAG_MAX_WRITE_BEHIND 6 /* 2 */ /* Only with raid1! */ +#define __CTR_FLAG_WRITE_MOSTLY 7 /* 2 */ /* Only with raid1! */ +#define __CTR_FLAG_STRIPE_CACHE 8 /* 2 */ /* Only with raid4/5/6! */ +#define __CTR_FLAG_REGION_SIZE 9 /* 2 */ /* Not with raid0! */ +#define __CTR_FLAG_RAID10_COPIES 10 /* 2 */ /* Only with raid10 */ +#define __CTR_FLAG_RAID10_FORMAT 11 /* 2 */ /* Only with raid10 */ +/* New for v1.9.0 */ +#define __CTR_FLAG_DELTA_DISKS 12 /* 2 */ /* Only with reshapable raid1/4/5/6/10! */ +#define __CTR_FLAG_DATA_OFFSET 13 /* 2 */ /* Only with reshapable raid4/5/6/10! */ +#define __CTR_FLAG_RAID10_USE_NEAR_SETS 14 /* 2 */ /* Only with raid10! */ + +/* * Flags for rs->ctr_flags field. */ -#define CTR_FLAG_SYNC 0x1 -#define CTR_FLAG_NOSYNC 0x2 -#define CTR_FLAG_REBUILD 0x4 -#define CTR_FLAG_DAEMON_SLEEP 0x8 -#define CTR_FLAG_MIN_RECOVERY_RATE 0x10 -#define CTR_FLAG_MAX_RECOVERY_RATE 0x20 -#define CTR_FLAG_MAX_WRITE_BEHIND 0x40 -#define CTR_FLAG_STRIPE_CACHE 0x80 -#define CTR_FLAG_REGION_SIZE 0x100 -#define CTR_FLAG_RAID10_COPIES 0x200 -#define CTR_FLAG_RAID10_FORMAT 0x400 +#define CTR_FLAG_SYNC (1 << __CTR_FLAG_SYNC) +#define CTR_FLAG_NOSYNC (1 << __CTR_FLAG_NOSYNC) +#define CTR_FLAG_REBUILD (1 << __CTR_FLAG_REBUILD) +#define CTR_FLAG_DAEMON_SLEEP (1 << __CTR_FLAG_DAEMON_SLEEP) +#define CTR_FLAG_MIN_RECOVERY_RATE (1 << __CTR_FLAG_MIN_RECOVERY_RATE) +#define CTR_FLAG_MAX_RECOVERY_RATE (1 << __CTR_FLAG_MAX_RECOVERY_RATE) +#define CTR_FLAG_MAX_WRITE_BEHIND (1 << __CTR_FLAG_MAX_WRITE_BEHIND) +#define CTR_FLAG_WRITE_MOSTLY (1 << __CTR_FLAG_WRITE_MOSTLY) +#define CTR_FLAG_STRIPE_CACHE (1 << __CTR_FLAG_STRIPE_CACHE) +#define CTR_FLAG_REGION_SIZE (1 << __CTR_FLAG_REGION_SIZE) +#define CTR_FLAG_RAID10_COPIES (1 << __CTR_FLAG_RAID10_COPIES) +#define CTR_FLAG_RAID10_FORMAT (1 << __CTR_FLAG_RAID10_FORMAT) +#define CTR_FLAG_DELTA_DISKS (1 << __CTR_FLAG_DELTA_DISKS) +#define CTR_FLAG_DATA_OFFSET (1 << __CTR_FLAG_DATA_OFFSET) +#define CTR_FLAG_RAID10_USE_NEAR_SETS (1 << __CTR_FLAG_RAID10_USE_NEAR_SETS) + +/* + * Definitions of various constructor flags to + * be used in checks of valid / invalid flags + * per raid level. + */ +/* Define all any sync flags */ +#define CTR_FLAGS_ANY_SYNC (CTR_FLAG_SYNC | CTR_FLAG_NOSYNC) + +/* Define flags for options without argument (e.g. 'nosync') */ +#define CTR_FLAG_OPTIONS_NO_ARGS (CTR_FLAGS_ANY_SYNC | \ + CTR_FLAG_RAID10_USE_NEAR_SETS) + +/* Define flags for options with one argument (e.g. 'delta_disks +2') */ +#define CTR_FLAG_OPTIONS_ONE_ARG (CTR_FLAG_REBUILD | \ + CTR_FLAG_WRITE_MOSTLY | \ + CTR_FLAG_DAEMON_SLEEP | \ + CTR_FLAG_MIN_RECOVERY_RATE | \ + CTR_FLAG_MAX_RECOVERY_RATE | \ + CTR_FLAG_MAX_WRITE_BEHIND | \ + CTR_FLAG_STRIPE_CACHE | \ + CTR_FLAG_REGION_SIZE | \ + CTR_FLAG_RAID10_COPIES | \ + CTR_FLAG_RAID10_FORMAT | \ + CTR_FLAG_DELTA_DISKS | \ + CTR_FLAG_DATA_OFFSET) + +/* Valid options definitions per raid level... */ + +/* "raid0" does only accept data offset */ +#define RAID0_VALID_FLAGS (CTR_FLAG_DATA_OFFSET) + +/* "raid1" does not accept stripe cache, data offset, delta_disks or any raid10 options */ +#define RAID1_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ + CTR_FLAG_REBUILD | \ + CTR_FLAG_WRITE_MOSTLY | \ + CTR_FLAG_DAEMON_SLEEP | \ + CTR_FLAG_MIN_RECOVERY_RATE | \ + CTR_FLAG_MAX_RECOVERY_RATE | \ + CTR_FLAG_MAX_WRITE_BEHIND | \ + CTR_FLAG_REGION_SIZE | \ + CTR_FLAG_DELTA_DISKS | \ + CTR_FLAG_DATA_OFFSET) + +/* "raid10" does not accept any raid1 or stripe cache options */ +#define RAID10_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ + CTR_FLAG_REBUILD | \ + CTR_FLAG_DAEMON_SLEEP | \ + CTR_FLAG_MIN_RECOVERY_RATE | \ + CTR_FLAG_MAX_RECOVERY_RATE | \ + CTR_FLAG_REGION_SIZE | \ + CTR_FLAG_RAID10_COPIES | \ + CTR_FLAG_RAID10_FORMAT | \ + CTR_FLAG_DELTA_DISKS | \ + CTR_FLAG_DATA_OFFSET | \ + CTR_FLAG_RAID10_USE_NEAR_SETS) + +/* + * "raid4/5/6" do not accept any raid1 or raid10 specific options + * + * "raid6" does not accept "nosync", because it is not guaranteed + * that both parity and q-syndrome are being written properly with + * any writes + */ +#define RAID45_VALID_FLAGS (CTR_FLAGS_ANY_SYNC | \ + CTR_FLAG_REBUILD | \ + CTR_FLAG_DAEMON_SLEEP | \ + CTR_FLAG_MIN_RECOVERY_RATE | \ + CTR_FLAG_MAX_RECOVERY_RATE | \ + CTR_FLAG_MAX_WRITE_BEHIND | \ + CTR_FLAG_STRIPE_CACHE | \ + CTR_FLAG_REGION_SIZE | \ + CTR_FLAG_DELTA_DISKS | \ + CTR_FLAG_DATA_OFFSET) + +#define RAID6_VALID_FLAGS (CTR_FLAG_SYNC | \ + CTR_FLAG_REBUILD | \ + CTR_FLAG_DAEMON_SLEEP | \ + CTR_FLAG_MIN_RECOVERY_RATE | \ + CTR_FLAG_MAX_RECOVERY_RATE | \ + CTR_FLAG_MAX_WRITE_BEHIND | \ + CTR_FLAG_STRIPE_CACHE | \ + CTR_FLAG_REGION_SIZE | \ + CTR_FLAG_DELTA_DISKS | \ + CTR_FLAG_DATA_OFFSET) +/* ...valid options definitions per raid level */ + +/* + * Flags for rs->runtime_flags field + * (RT_FLAG prefix meaning "runtime flag") + * + * These are all internal and used to define runtime state, + * e.g. to prevent another resume from preresume processing + * the raid set all over again. + */ +#define RT_FLAG_RS_PRERESUMED 0 +#define RT_FLAG_RS_RESUMED 1 +#define RT_FLAG_RS_BITMAP_LOADED 2 +#define RT_FLAG_UPDATE_SBS 3 +#define RT_FLAG_RESHAPE_RS 4 +#define RT_FLAG_KEEP_RS_FROZEN 5 + +/* 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) + +/* + * raid set level, layout and chunk sectors backup/restore + */ +struct rs_layout { + int new_level; + int new_layout; + int new_chunk_sectors; +}; struct raid_set { struct dm_target *ti; uint32_t bitmap_loaded; - uint32_t ctr_flags; + uint32_t stripe_cache_entries; + unsigned long ctr_flags; + unsigned long runtime_flags; + + uint64_t rebuild_disks[DISKS_ARRAY_ELEMS]; + + int raid_disks; + int delta_disks; + int data_offset; + int raid10_copies; + int requested_bitmap_chunk_sectors; struct mddev md; struct raid_type *raid_type; @@ -73,82 +228,446 @@ struct raid_set { struct raid_dev dev[0]; }; +static void rs_config_backup(struct raid_set *rs, struct rs_layout *l) +{ + struct mddev *mddev = &rs->md; + + l->new_level = mddev->new_level; + l->new_layout = mddev->new_layout; + l->new_chunk_sectors = mddev->new_chunk_sectors; +} + +static void rs_config_restore(struct raid_set *rs, struct rs_layout *l) +{ + struct mddev *mddev = &rs->md; + + mddev->new_level = l->new_level; + mddev->new_layout = l->new_layout; + mddev->new_chunk_sectors = l->new_chunk_sectors; +} + +/* raid10 algorithms (i.e. formats) */ +#define ALGORITHM_RAID10_DEFAULT 0 +#define ALGORITHM_RAID10_NEAR 1 +#define ALGORITHM_RAID10_OFFSET 2 +#define ALGORITHM_RAID10_FAR 3 + /* Supported raid types and properties. */ static struct raid_type { const char *name; /* RAID algorithm. */ const char *descr; /* Descriptor text for logging. */ - const unsigned parity_devs; /* # of parity devices. */ - const unsigned minimal_devs; /* minimal # of devices in set. */ - const unsigned level; /* RAID level. */ - const unsigned algorithm; /* RAID algorithm. */ + const unsigned int parity_devs; /* # of parity devices. */ + const unsigned int minimal_devs;/* minimal # of devices in set. */ + const unsigned int level; /* RAID level. */ + const unsigned int algorithm; /* RAID algorithm. */ } raid_types[] = { - {"raid0", "RAID0 (striping)", 0, 2, 0, 0 /* NONE */}, - {"raid1", "RAID1 (mirroring)", 0, 2, 1, 0 /* NONE */}, - {"raid10", "RAID10 (striped mirrors)", 0, 2, 10, UINT_MAX /* Varies */}, - {"raid4", "RAID4 (dedicated parity disk)", 1, 2, 5, ALGORITHM_PARITY_0}, - {"raid5_la", "RAID5 (left asymmetric)", 1, 2, 5, ALGORITHM_LEFT_ASYMMETRIC}, - {"raid5_ra", "RAID5 (right asymmetric)", 1, 2, 5, ALGORITHM_RIGHT_ASYMMETRIC}, - {"raid5_ls", "RAID5 (left symmetric)", 1, 2, 5, ALGORITHM_LEFT_SYMMETRIC}, - {"raid5_rs", "RAID5 (right symmetric)", 1, 2, 5, ALGORITHM_RIGHT_SYMMETRIC}, - {"raid6_zr", "RAID6 (zero restart)", 2, 4, 6, ALGORITHM_ROTATING_ZERO_RESTART}, - {"raid6_nr", "RAID6 (N restart)", 2, 4, 6, ALGORITHM_ROTATING_N_RESTART}, - {"raid6_nc", "RAID6 (N continue)", 2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE} + {"raid0", "raid0 (striping)", 0, 2, 0, 0 /* NONE */}, + {"raid1", "raid1 (mirroring)", 0, 2, 1, 0 /* NONE */}, + {"raid10_far", "raid10 far (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_FAR}, + {"raid10_offset", "raid10 offset (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_OFFSET}, + {"raid10_near", "raid10 near (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_NEAR}, + {"raid10", "raid10 (striped mirrors)", 0, 2, 10, ALGORITHM_RAID10_DEFAULT}, + {"raid4", "raid4 (dedicated last parity disk)", 1, 2, 4, ALGORITHM_PARITY_N}, /* raid4 layout = raid5_n */ + {"raid5_n", "raid5 (dedicated last parity disk)", 1, 2, 5, ALGORITHM_PARITY_N}, + {"raid5_ls", "raid5 (left symmetric)", 1, 2, 5, ALGORITHM_LEFT_SYMMETRIC}, + {"raid5_rs", "raid5 (right symmetric)", 1, 2, 5, ALGORITHM_RIGHT_SYMMETRIC}, + {"raid5_la", "raid5 (left asymmetric)", 1, 2, 5, ALGORITHM_LEFT_ASYMMETRIC}, + {"raid5_ra", "raid5 (right asymmetric)", 1, 2, 5, ALGORITHM_RIGHT_ASYMMETRIC}, + {"raid6_zr", "raid6 (zero restart)", 2, 4, 6, ALGORITHM_ROTATING_ZERO_RESTART}, + {"raid6_nr", "raid6 (N restart)", 2, 4, 6, ALGORITHM_ROTATING_N_RESTART}, + {"raid6_nc", "raid6 (N continue)", 2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE}, + {"raid6_n_6", "raid6 (dedicated parity/Q n/6)", 2, 4, 6, ALGORITHM_PARITY_N_6}, + {"raid6_ls_6", "raid6 (left symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_SYMMETRIC_6}, + {"raid6_rs_6", "raid6 (right symmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_SYMMETRIC_6}, + {"raid6_la_6", "raid6 (left asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_LEFT_ASYMMETRIC_6}, + {"raid6_ra_6", "raid6 (right asymmetric dedicated Q 6)", 2, 4, 6, ALGORITHM_RIGHT_ASYMMETRIC_6} +}; + +/* True, if @v is in inclusive range [@min, @max] */ +static bool __within_range(long v, long min, long max) +{ + return v >= min && v <= max; +} + +/* All table line arguments are defined here */ +static struct arg_name_flag { + const unsigned long flag; + const char *name; +} __arg_name_flags[] = { + { CTR_FLAG_SYNC, "sync"}, + { CTR_FLAG_NOSYNC, "nosync"}, + { CTR_FLAG_REBUILD, "rebuild"}, + { CTR_FLAG_DAEMON_SLEEP, "daemon_sleep"}, + { CTR_FLAG_MIN_RECOVERY_RATE, "min_recovery_rate"}, + { CTR_FLAG_MAX_RECOVERY_RATE, "max_recovery_rate"}, + { CTR_FLAG_MAX_WRITE_BEHIND, "max_write_behind"}, + { CTR_FLAG_WRITE_MOSTLY, "write_mostly"}, + { CTR_FLAG_STRIPE_CACHE, "stripe_cache"}, + { CTR_FLAG_REGION_SIZE, "region_size"}, + { CTR_FLAG_RAID10_COPIES, "raid10_copies"}, + { CTR_FLAG_RAID10_FORMAT, "raid10_format"}, + { CTR_FLAG_DATA_OFFSET, "data_offset"}, + { CTR_FLAG_DELTA_DISKS, "delta_disks"}, + { CTR_FLAG_RAID10_USE_NEAR_SETS, "raid10_use_near_sets"}, }; -static char *raid10_md_layout_to_format(int layout) +/* Return argument name string for given @flag */ +static const char *dm_raid_arg_name_by_flag(const uint32_t flag) +{ + if (hweight32(flag) == 1) { + struct arg_name_flag *anf = __arg_name_flags + ARRAY_SIZE(__arg_name_flags); + + while (anf-- > __arg_name_flags) + if (flag & anf->flag) + return anf->name; + + } else + DMERR("%s called with more than one flag!", __func__); + + return NULL; +} + +/* + * Bool helpers to test for various raid levels of a raid set. + * It's level as reported by the superblock rather than + * the requested raid_type passed to the constructor. + */ +/* Return true, if raid set in @rs is raid0 */ +static bool rs_is_raid0(struct raid_set *rs) +{ + return !rs->md.level; +} + +/* Return true, if raid set in @rs is raid1 */ +static bool rs_is_raid1(struct raid_set *rs) +{ + return rs->md.level == 1; +} + +/* Return true, if raid set in @rs is raid10 */ +static bool rs_is_raid10(struct raid_set *rs) +{ + return rs->md.level == 10; +} + +/* Return true, if raid set in @rs is level 6 */ +static bool rs_is_raid6(struct raid_set *rs) +{ + return rs->md.level == 6; +} + +/* Return true, if raid set in @rs is level 4, 5 or 6 */ +static bool rs_is_raid456(struct raid_set *rs) +{ + return __within_range(rs->md.level, 4, 6); +} + +/* Return true, if raid set in @rs is reshapable */ +static bool __is_raid10_far(int layout); +static bool rs_is_reshapable(struct raid_set *rs) +{ + return rs_is_raid456(rs) || + (rs_is_raid10(rs) && !__is_raid10_far(rs->md.new_layout)); +} + +/* Return true, if raid set in @rs is recovering */ +static bool rs_is_recovering(struct raid_set *rs) +{ + return rs->md.recovery_cp < rs->dev[0].rdev.sectors; +} + +/* Return true, if raid set in @rs is reshaping */ +static bool rs_is_reshaping(struct raid_set *rs) +{ + return rs->md.reshape_position != MaxSector; +} + +/* + * bool helpers to test for various raid levels of a raid type @rt + */ + +/* Return true, if raid type in @rt is raid0 */ +static bool rt_is_raid0(struct raid_type *rt) +{ + return !rt->level; +} + +/* Return true, if raid type in @rt is raid1 */ +static bool rt_is_raid1(struct raid_type *rt) +{ + return rt->level == 1; +} + +/* Return true, if raid type in @rt is raid10 */ +static bool rt_is_raid10(struct raid_type *rt) +{ + return rt->level == 10; +} + +/* Return true, if raid type in @rt is raid4/5 */ +static bool rt_is_raid45(struct raid_type *rt) +{ + return __within_range(rt->level, 4, 5); +} + +/* Return true, if raid type in @rt is raid6 */ +static bool rt_is_raid6(struct raid_type *rt) +{ + return rt->level == 6; +} + +/* Return true, if raid type in @rt is raid4/5/6 */ +static bool rt_is_raid456(struct raid_type *rt) +{ + return __within_range(rt->level, 4, 6); +} +/* END: raid level bools */ + +/* Return valid ctr flags for the raid level of @rs */ +static unsigned long __valid_flags(struct raid_set *rs) +{ + if (rt_is_raid0(rs->raid_type)) + return RAID0_VALID_FLAGS; + else if (rt_is_raid1(rs->raid_type)) + return RAID1_VALID_FLAGS; + else if (rt_is_raid10(rs->raid_type)) + return RAID10_VALID_FLAGS; + else if (rt_is_raid45(rs->raid_type)) + return RAID45_VALID_FLAGS; + else if (rt_is_raid6(rs->raid_type)) + return RAID6_VALID_FLAGS; + + return 0; +} + +/* + * Check for valid flags set on @rs + * + * Has to be called after parsing of the ctr flags! + */ +static int rs_check_for_valid_flags(struct raid_set *rs) +{ + if (rs->ctr_flags & ~__valid_flags(rs)) { + rs->ti->error = "Invalid flags combination"; + return -EINVAL; + } + + return 0; +} + +/* MD raid10 bit definitions and helpers */ +#define RAID10_OFFSET (1 << 16) /* stripes with data copies area adjacent on devices */ +#define RAID10_BROCKEN_USE_FAR_SETS (1 << 17) /* Broken in raid10.c: use sets instead of whole stripe rotation */ +#define RAID10_USE_FAR_SETS (1 << 18) /* Use sets instead of whole stripe rotation */ +#define RAID10_FAR_COPIES_SHIFT 8 /* raid10 # far copies shift (2nd byte of layout) */ + +/* Return md raid10 near copies for @layout */ +static unsigned int __raid10_near_copies(int layout) +{ + return layout & 0xFF; +} + +/* Return md raid10 far copies for @layout */ +static unsigned int __raid10_far_copies(int layout) +{ + return __raid10_near_copies(layout >> RAID10_FAR_COPIES_SHIFT); +} + +/* Return true if md raid10 offset for @layout */ +static bool __is_raid10_offset(int layout) +{ + return !!(layout & RAID10_OFFSET); +} + +/* Return true if md raid10 near for @layout */ +static bool __is_raid10_near(int layout) +{ + return !__is_raid10_offset(layout) && __raid10_near_copies(layout) > 1; +} + +/* Return true if md raid10 far for @layout */ +static bool __is_raid10_far(int layout) +{ + return !__is_raid10_offset(layout) && __raid10_far_copies(layout) > 1; +} + +/* Return md raid10 layout string for @layout */ +static const char *raid10_md_layout_to_format(int layout) { /* - * Bit 16 and 17 stand for "offset" and "use_far_sets" + * Bit 16 stands for "offset" + * (i.e. adjacent stripes hold copies) + * * Refer to MD's raid10.c for details */ - if ((layout & 0x10000) && (layout & 0x20000)) + if (__is_raid10_offset(layout)) return "offset"; - if ((layout & 0xFF) > 1) + if (__raid10_near_copies(layout) > 1) return "near"; + WARN_ON(__raid10_far_copies(layout) < 2); + return "far"; } -static unsigned raid10_md_layout_to_copies(int layout) +/* Return md raid10 algorithm for @name */ +static int raid10_name_to_format(const char *name) { - if ((layout & 0xFF) > 1) - return layout & 0xFF; - return (layout >> 8) & 0xFF; + if (!strcasecmp(name, "near")) + return ALGORITHM_RAID10_NEAR; + else if (!strcasecmp(name, "offset")) + return ALGORITHM_RAID10_OFFSET; + else if (!strcasecmp(name, "far")) + return ALGORITHM_RAID10_FAR; + + return -EINVAL; } -static int raid10_format_to_md_layout(char *format, unsigned copies) +/* Return md raid10 copies for @layout */ +static unsigned int raid10_md_layout_to_copies(int layout) { - unsigned n = 1, f = 1; + return max(__raid10_near_copies(layout), __raid10_far_copies(layout)); +} - if (!strcasecmp("near", format)) +/* Return md raid10 format id for @format string */ +static int raid10_format_to_md_layout(struct raid_set *rs, + unsigned int algorithm, + unsigned int copies) +{ + unsigned int n = 1, f = 1, r = 0; + + /* + * MD resilienece flaw: + * + * enabling use_far_sets for far/offset formats causes copies + * to be colocated on the same devs together with their origins! + * + * -> disable it for now in the definition above + */ + if (algorithm == ALGORITHM_RAID10_DEFAULT || + algorithm == ALGORITHM_RAID10_NEAR) n = copies; - else + + else if (algorithm == ALGORITHM_RAID10_OFFSET) { f = copies; + r = RAID10_OFFSET; + if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) + r |= RAID10_USE_FAR_SETS; - if (!strcasecmp("offset", format)) - return 0x30000 | (f << 8) | n; + } else if (algorithm == ALGORITHM_RAID10_FAR) { + f = copies; + r = !RAID10_OFFSET; + if (!test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) + r |= RAID10_USE_FAR_SETS; - if (!strcasecmp("far", format)) - return 0x20000 | (f << 8) | n; + } else + return -EINVAL; - return (f << 8) | n; + return r | (f << RAID10_FAR_COPIES_SHIFT) | n; } +/* END: MD raid10 bit definitions and helpers */ -static struct raid_type *get_raid_type(char *name) +/* Check for any of the raid10 algorithms */ +static bool __got_raid10(struct raid_type *rtp, const int layout) { - int i; + if (rtp->level == 10) { + switch (rtp->algorithm) { + case ALGORITHM_RAID10_DEFAULT: + case ALGORITHM_RAID10_NEAR: + return __is_raid10_near(layout); + case ALGORITHM_RAID10_OFFSET: + return __is_raid10_offset(layout); + case ALGORITHM_RAID10_FAR: + return __is_raid10_far(layout); + default: + break; + } + } + + return false; +} - for (i = 0; i < ARRAY_SIZE(raid_types); i++) - if (!strcmp(raid_types[i].name, name)) - return &raid_types[i]; +/* Return raid_type for @name */ +static struct raid_type *get_raid_type(const char *name) +{ + struct raid_type *rtp = raid_types + ARRAY_SIZE(raid_types); + + while (rtp-- > raid_types) + if (!strcasecmp(rtp->name, name)) + return rtp; return NULL; } -static struct raid_set *context_alloc(struct dm_target *ti, struct raid_type *raid_type, unsigned raid_devs) +/* Return raid_type for @name based derived from @level and @layout */ +static struct raid_type *get_raid_type_by_ll(const int level, const int layout) { - unsigned i; + struct raid_type *rtp = raid_types + ARRAY_SIZE(raid_types); + + while (rtp-- > raid_types) { + /* RAID10 special checks based on @layout flags/properties */ + if (rtp->level == level && + (__got_raid10(rtp, layout) || rtp->algorithm == layout)) + return rtp; + } + + 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) +{ + 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 + * is unintended in case of out-of-place reshaping + */ + rdev_for_each(rdev, mddev) + rdev->sectors = mddev->dev_sectors; + + set_capacity(gendisk, mddev->array_sectors); + revalidate_disk(gendisk); +} + +/* + * Set the mddev properties in @rs to the current + * ones retrieved from the freshest superblock + */ +static void rs_set_cur(struct raid_set *rs) +{ + struct mddev *mddev = &rs->md; + + mddev->new_level = mddev->level; + mddev->new_layout = mddev->layout; + mddev->new_chunk_sectors = mddev->chunk_sectors; +} + +/* + * Set the mddev properties in @rs to the new + * ones requested by the ctr + */ +static void rs_set_new(struct raid_set *rs) +{ + struct mddev *mddev = &rs->md; + + mddev->level = mddev->new_level; + mddev->layout = mddev->new_layout; + mddev->chunk_sectors = mddev->new_chunk_sectors; + mddev->raid_disks = rs->raid_disks; + mddev->delta_disks = 0; +} + +static struct raid_set *raid_set_alloc(struct dm_target *ti, struct raid_type *raid_type, + unsigned int raid_devs) +{ + unsigned int i; struct raid_set *rs; if (raid_devs <= raid_type->parity_devs) { @@ -164,15 +683,19 @@ static struct raid_set *context_alloc(struct dm_target *ti, struct raid_type *ra mddev_init(&rs->md); + rs->raid_disks = raid_devs; + rs->delta_disks = 0; + rs->ti = ti; rs->raid_type = raid_type; + rs->stripe_cache_entries = 256; rs->md.raid_disks = raid_devs; rs->md.level = raid_type->level; rs->md.new_level = rs->md.level; rs->md.layout = raid_type->algorithm; rs->md.new_layout = rs->md.layout; rs->md.delta_disks = 0; - rs->md.recovery_cp = 0; + rs->md.recovery_cp = MaxSector; for (i = 0; i < raid_devs; i++) md_rdev_init(&rs->dev[i].rdev); @@ -189,11 +712,11 @@ static struct raid_set *context_alloc(struct dm_target *ti, struct raid_type *ra return rs; } -static void context_free(struct raid_set *rs) +static void raid_set_free(struct raid_set *rs) { int i; - for (i = 0; i < rs->md.raid_disks; i++) { + for (i = 0; i < rs->raid_disks; i++) { if (rs->dev[i].meta_dev) dm_put_device(rs->ti, rs->dev[i].meta_dev); md_rdev_clear(&rs->dev[i].rdev); @@ -218,16 +741,22 @@ static void context_free(struct raid_set *rs) * <meta_dev> - * * This code parses those words. If there is a failure, - * the caller must use context_free to unwind the operations. + * the caller must use raid_set_free() to unwind the operations. */ -static int dev_parms(struct raid_set *rs, char **argv) +static int parse_dev_params(struct raid_set *rs, struct dm_arg_set *as) { int i; int rebuild = 0; int metadata_available = 0; - int ret = 0; + int r = 0; + const char *arg; - for (i = 0; i < rs->md.raid_disks; i++, argv += 2) { + /* Put off the number of raid devices argument to get to dev pairs */ + arg = dm_shift_arg(as); + if (!arg) + return -EINVAL; + + for (i = 0; i < rs->raid_disks; i++) { rs->dev[i].rdev.raid_disk = i; rs->dev[i].meta_dev = NULL; @@ -240,39 +769,49 @@ static int dev_parms(struct raid_set *rs, char **argv) rs->dev[i].rdev.data_offset = 0; rs->dev[i].rdev.mddev = &rs->md; - if (strcmp(argv[0], "-")) { - ret = dm_get_device(rs->ti, argv[0], - dm_table_get_mode(rs->ti->table), - &rs->dev[i].meta_dev); - rs->ti->error = "RAID metadata device lookup failure"; - if (ret) - return ret; + arg = dm_shift_arg(as); + if (!arg) + return -EINVAL; + + if (strcmp(arg, "-")) { + r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), + &rs->dev[i].meta_dev); + if (r) { + rs->ti->error = "RAID metadata device lookup failure"; + return r; + } rs->dev[i].rdev.sb_page = alloc_page(GFP_KERNEL); - if (!rs->dev[i].rdev.sb_page) + if (!rs->dev[i].rdev.sb_page) { + rs->ti->error = "Failed to allocate superblock page"; return -ENOMEM; + } } - if (!strcmp(argv[1], "-")) { + arg = dm_shift_arg(as); + if (!arg) + return -EINVAL; + + if (!strcmp(arg, "-")) { if (!test_bit(In_sync, &rs->dev[i].rdev.flags) && (!rs->dev[i].rdev.recovery_offset)) { rs->ti->error = "Drive designated for rebuild not specified"; return -EINVAL; } - rs->ti->error = "No data device supplied with metadata device"; - if (rs->dev[i].meta_dev) + if (rs->dev[i].meta_dev) { + rs->ti->error = "No data device supplied with metadata device"; return -EINVAL; + } continue; } - ret = dm_get_device(rs->ti, argv[1], - dm_table_get_mode(rs->ti->table), - &rs->dev[i].data_dev); - if (ret) { + r = dm_get_device(rs->ti, arg, dm_table_get_mode(rs->ti->table), + &rs->dev[i].data_dev); + if (r) { rs->ti->error = "RAID device lookup failure"; - return ret; + return r; } if (rs->dev[i].meta_dev) { @@ -280,7 +819,7 @@ static int dev_parms(struct raid_set *rs, char **argv) rs->dev[i].rdev.meta_bdev = rs->dev[i].meta_dev->bdev; } rs->dev[i].rdev.bdev = rs->dev[i].data_dev->bdev; - list_add(&rs->dev[i].rdev.same_set, &rs->md.disks); + list_add_tail(&rs->dev[i].rdev.same_set, &rs->md.disks); if (!test_bit(In_sync, &rs->dev[i].rdev.flags)) rebuild++; } @@ -301,8 +840,7 @@ static int dev_parms(struct raid_set *rs, char **argv) * * User could specify 'nosync' option if desperate. */ - DMERR("Unable to rebuild drive while array is not in-sync"); - rs->ti->error = "RAID device lookup failure"; + rs->ti->error = "Unable to rebuild drive while array is not in-sync"; return -EINVAL; } @@ -325,7 +863,7 @@ static int validate_region_size(struct raid_set *rs, unsigned long region_size) if (!region_size) { /* - * Choose a reasonable default. All figures in sectors. + * Choose a reasonable default. All figures in sectors. */ if (min_region_size > (1 << 13)) { /* If not a power of 2, make it the next power of 2 */ @@ -366,7 +904,7 @@ static int validate_region_size(struct raid_set *rs, unsigned long region_size) /* * Convert sectors to bytes. */ - rs->md.bitmap_info.chunksize = (region_size << 9); + rs->md.bitmap_info.chunksize = to_bytes(region_size); return 0; } @@ -382,9 +920,9 @@ static int validate_region_size(struct raid_set *rs, unsigned long region_size) */ static int validate_raid_redundancy(struct raid_set *rs) { - unsigned i, rebuild_cnt = 0; - unsigned rebuilds_per_group = 0, copies, d; - unsigned group_size, last_group_start; + unsigned int i, rebuild_cnt = 0; + unsigned int rebuilds_per_group = 0, copies; + unsigned int group_size, last_group_start; for (i = 0; i < rs->md.raid_disks; i++) if (!test_bit(In_sync, &rs->dev[i].rdev.flags) || @@ -403,7 +941,7 @@ static int validate_raid_redundancy(struct raid_set *rs) goto too_many; break; case 10: - copies = raid10_md_layout_to_copies(rs->md.layout); + copies = raid10_md_layout_to_copies(rs->md.new_layout); if (rebuild_cnt < copies) break; @@ -417,17 +955,16 @@ static int validate_raid_redundancy(struct raid_set *rs) * simple case where the number of devices is a multiple of the * number of copies, we must also handle cases where the number * of devices is not a multiple of the number of copies. - * E.g. dev1 dev2 dev3 dev4 dev5 - * A A B B C - * C D D E E + * E.g. dev1 dev2 dev3 dev4 dev5 + * A A B B C + * C D D E E */ - if (!strcmp("near", raid10_md_layout_to_format(rs->md.layout))) { - for (i = 0; i < rs->md.raid_disks * copies; i++) { + if (__is_raid10_near(rs->md.new_layout)) { + for (i = 0; i < rs->md.raid_disks; i++) { if (!(i % copies)) rebuilds_per_group = 0; - d = i % rs->md.raid_disks; - if ((!rs->dev[d].rdev.sb_page || - !test_bit(In_sync, &rs->dev[d].rdev.flags)) && + if ((!rs->dev[i].rdev.sb_page || + !test_bit(In_sync, &rs->dev[i].rdev.flags)) && (++rebuilds_per_group >= copies)) goto too_many; } @@ -442,7 +979,7 @@ static int validate_raid_redundancy(struct raid_set *rs) * use the 'use_far_sets' variant.) * * This check is somewhat complicated by the need to account - * for arrays that are not a multiple of (far) copies. This + * for arrays that are not a multiple of (far) copies. This * results in the need to treat the last (potentially larger) * set differently. */ @@ -475,42 +1012,48 @@ too_many: * * Argument definitions * <chunk_size> The number of sectors per disk that - * will form the "stripe" + * will form the "stripe" * [[no]sync] Force or prevent recovery of the - * entire array + * entire array * [rebuild <idx>] Rebuild the drive indicated by the index * [daemon_sleep <ms>] Time between bitmap daemon work to - * clear bits + * clear bits * [min_recovery_rate <kB/sec/disk>] Throttle RAID initialization * [max_recovery_rate <kB/sec/disk>] Throttle RAID initialization * [write_mostly <idx>] Indicate a write mostly drive via index * [max_write_behind <sectors>] See '-write-behind=' (man mdadm) * [stripe_cache <sectors>] Stripe cache size for higher RAIDs - * [region_size <sectors>] Defines granularity of bitmap + * [region_size <sectors>] Defines granularity of bitmap * * RAID10-only options: - * [raid10_copies <# copies>] Number of copies. (Default: 2) + * [raid10_copies <# copies>] Number of copies. (Default: 2) * [raid10_format <near|far|offset>] Layout algorithm. (Default: near) */ -static int parse_raid_params(struct raid_set *rs, char **argv, - unsigned num_raid_params) -{ - char *raid10_format = "near"; - unsigned raid10_copies = 2; - unsigned i; - unsigned long value, region_size = 0; - sector_t sectors_per_dev = rs->ti->len; +static int parse_raid_params(struct raid_set *rs, struct dm_arg_set *as, + unsigned int num_raid_params) +{ + int value, raid10_format = ALGORITHM_RAID10_DEFAULT; + unsigned int raid10_copies = 2; + unsigned int i, write_mostly = 0; + unsigned int region_size = 0; sector_t max_io_len; - char *key; + const char *arg, *key; + struct raid_dev *rd; + struct raid_type *rt = rs->raid_type; + + arg = dm_shift_arg(as); + num_raid_params--; /* Account for chunk_size argument */ + + if (kstrtoint(arg, 10, &value) < 0) { + rs->ti->error = "Bad numerical argument given for chunk_size"; + return -EINVAL; + } /* * First, parse the in-order required arguments * "chunk_size" is the only argument of this type. */ - if ((kstrtoul(argv[0], 10, &value) < 0)) { - rs->ti->error = "Bad chunk size"; - return -EINVAL; - } else if (rs->raid_type->level == 1) { + if (rt_is_raid1(rt)) { if (value) DMERR("Ignoring chunk size parameter for RAID 1"); value = 0; @@ -523,8 +1066,6 @@ static int parse_raid_params(struct raid_set *rs, char **argv, } rs->md.new_chunk_sectors = rs->md.chunk_sectors = value; - argv++; - num_raid_params--; /* * We set each individual device as In_sync with a completed @@ -532,18 +1073,18 @@ static int parse_raid_params(struct raid_set *rs, char **argv, * replacement then one of the following cases applies: * * 1) User specifies 'rebuild'. - * - Device is reset when param is read. + * - Device is reset when param is read. * 2) A new device is supplied. - * - No matching superblock found, resets device. + * - No matching superblock found, resets device. * 3) Device failure was transient and returns on reload. - * - Failure noticed, resets device for bitmap replay. + * - Failure noticed, resets device for bitmap replay. * 4) Device hadn't completed recovery after previous failure. - * - Superblock is read and overrides recovery_offset. + * - Superblock is read and overrides recovery_offset. * * What is found in the superblocks of the devices is always * authoritative, unless 'rebuild' or '[no]sync' was specified. */ - for (i = 0; i < rs->md.raid_disks; i++) { + for (i = 0; i < rs->raid_disks; i++) { set_bit(In_sync, &rs->dev[i].rdev.flags); rs->dev[i].rdev.recovery_offset = MaxSector; } @@ -552,72 +1093,112 @@ static int parse_raid_params(struct raid_set *rs, char **argv, * Second, parse the unordered optional arguments */ for (i = 0; i < num_raid_params; i++) { - if (!strcasecmp(argv[i], "nosync")) { - rs->md.recovery_cp = MaxSector; - rs->ctr_flags |= CTR_FLAG_NOSYNC; + key = dm_shift_arg(as); + if (!key) { + rs->ti->error = "Not enough raid parameters given"; + return -EINVAL; + } + + if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC))) { + if (test_and_set_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { + rs->ti->error = "Only one 'nosync' argument allowed"; + return -EINVAL; + } continue; } - if (!strcasecmp(argv[i], "sync")) { - rs->md.recovery_cp = 0; - rs->ctr_flags |= CTR_FLAG_SYNC; + if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_SYNC))) { + if (test_and_set_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) { + rs->ti->error = "Only one 'sync' argument allowed"; + return -EINVAL; + } + continue; + } + if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_USE_NEAR_SETS))) { + if (test_and_set_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { + rs->ti->error = "Only one 'raid10_use_new_sets' argument allowed"; + return -EINVAL; + } continue; } - /* The rest of the optional arguments come in key/value pairs */ - if ((i + 1) >= num_raid_params) { + arg = dm_shift_arg(as); + i++; /* Account for the argument pairs */ + if (!arg) { rs->ti->error = "Wrong number of raid parameters given"; return -EINVAL; } - key = argv[i++]; + /* + * Parameters that take a string value are checked here. + */ - /* Parameters that take a string value are checked here. */ - if (!strcasecmp(key, "raid10_format")) { - if (rs->raid_type->level != 10) { + if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT))) { + if (test_and_set_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) { + rs->ti->error = "Only one 'raid10_format' argument pair allowed"; + return -EINVAL; + } + if (!rt_is_raid10(rt)) { rs->ti->error = "'raid10_format' is an invalid parameter for this RAID type"; return -EINVAL; } - if (strcmp("near", argv[i]) && - strcmp("far", argv[i]) && - strcmp("offset", argv[i])) { + raid10_format = raid10_name_to_format(arg); + if (raid10_format < 0) { rs->ti->error = "Invalid 'raid10_format' value given"; - return -EINVAL; + return raid10_format; } - raid10_format = argv[i]; - rs->ctr_flags |= CTR_FLAG_RAID10_FORMAT; continue; } - if (kstrtoul(argv[i], 10, &value) < 0) { + if (kstrtoint(arg, 10, &value) < 0) { rs->ti->error = "Bad numerical argument given in raid params"; return -EINVAL; } - /* Parameters that take a numeric value are checked here */ - if (!strcasecmp(key, "rebuild")) { - if (value >= rs->md.raid_disks) { + if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD))) { + /* + * "rebuild" is being passed in by userspace to provide + * indexes of replaced devices and to set up additional + * devices on raid level takeover. + */ + if (!__within_range(value, 0, rs->raid_disks - 1)) { rs->ti->error = "Invalid rebuild index given"; return -EINVAL; } - clear_bit(In_sync, &rs->dev[value].rdev.flags); - rs->dev[value].rdev.recovery_offset = 0; - rs->ctr_flags |= CTR_FLAG_REBUILD; - } else if (!strcasecmp(key, "write_mostly")) { - if (rs->raid_type->level != 1) { + + if (test_and_set_bit(value, (void *) rs->rebuild_disks)) { + rs->ti->error = "rebuild for this index already given"; + return -EINVAL; + } + + rd = rs->dev + value; + clear_bit(In_sync, &rd->rdev.flags); + clear_bit(Faulty, &rd->rdev.flags); + rd->rdev.recovery_offset = 0; + set_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags); + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY))) { + if (!rt_is_raid1(rt)) { rs->ti->error = "write_mostly option is only valid for RAID1"; return -EINVAL; } - if (value >= rs->md.raid_disks) { - rs->ti->error = "Invalid write_mostly drive index given"; + + if (!__within_range(value, 0, rs->md.raid_disks - 1)) { + rs->ti->error = "Invalid write_mostly index given"; return -EINVAL; } + + write_mostly++; set_bit(WriteMostly, &rs->dev[value].rdev.flags); - } else if (!strcasecmp(key, "max_write_behind")) { - if (rs->raid_type->level != 1) { + set_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags); + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND))) { + if (!rt_is_raid1(rt)) { rs->ti->error = "max_write_behind option is only valid for RAID1"; return -EINVAL; } - rs->ctr_flags |= CTR_FLAG_MAX_WRITE_BEHIND; + + if (test_and_set_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) { + rs->ti->error = "Only one max_write_behind argument pair allowed"; + return -EINVAL; + } /* * In device-mapper, we specify things in sectors, but @@ -628,64 +1209,121 @@ static int parse_raid_params(struct raid_set *rs, char **argv, rs->ti->error = "Max write-behind limit out of range"; return -EINVAL; } + rs->md.bitmap_info.max_write_behind = value; - } else if (!strcasecmp(key, "daemon_sleep")) { - rs->ctr_flags |= CTR_FLAG_DAEMON_SLEEP; + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP))) { + if (test_and_set_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) { + rs->ti->error = "Only one daemon_sleep argument pair allowed"; + return -EINVAL; + } if (!value || (value > MAX_SCHEDULE_TIMEOUT)) { rs->ti->error = "daemon sleep period out of range"; return -EINVAL; } rs->md.bitmap_info.daemon_sleep = value; - } else if (!strcasecmp(key, "stripe_cache")) { - rs->ctr_flags |= CTR_FLAG_STRIPE_CACHE; + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET))) { + /* Userspace passes new data_offset after having extended the the data image LV */ + if (test_and_set_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { + rs->ti->error = "Only one data_offset argument pair allowed"; + return -EINVAL; + } + /* Ensure sensible data offset */ + if (value < 0 || + (value && (value < MIN_FREE_RESHAPE_SPACE || value % to_sector(PAGE_SIZE)))) { + rs->ti->error = "Bogus data_offset value"; + return -EINVAL; + } + rs->data_offset = value; + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS))) { + /* Define the +/-# of disks to add to/remove from the given raid set */ + if (test_and_set_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { + rs->ti->error = "Only one delta_disks argument pair allowed"; + return -EINVAL; + } + /* Ensure MAX_RAID_DEVICES and raid type minimal_devs! */ + if (!__within_range(abs(value), 1, MAX_RAID_DEVICES - rt->minimal_devs)) { + rs->ti->error = "Too many delta_disk requested"; + return -EINVAL; + } - /* - * In device-mapper, we specify things in sectors, but - * MD records this value in kB - */ - value /= 2; + rs->delta_disks = value; + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE))) { + if (test_and_set_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) { + rs->ti->error = "Only one stripe_cache argument pair allowed"; + return -EINVAL; + } - if ((rs->raid_type->level != 5) && - (rs->raid_type->level != 6)) { + if (!rt_is_raid456(rt)) { rs->ti->error = "Inappropriate argument: stripe_cache"; return -EINVAL; } - if (raid5_set_cache_size(&rs->md, (int)value)) { - rs->ti->error = "Bad stripe_cache size"; + + rs->stripe_cache_entries = value; + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE))) { + if (test_and_set_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) { + rs->ti->error = "Only one min_recovery_rate argument pair allowed"; return -EINVAL; } - } else if (!strcasecmp(key, "min_recovery_rate")) { - rs->ctr_flags |= CTR_FLAG_MIN_RECOVERY_RATE; if (value > INT_MAX) { rs->ti->error = "min_recovery_rate out of range"; return -EINVAL; } rs->md.sync_speed_min = (int)value; - } else if (!strcasecmp(key, "max_recovery_rate")) { - rs->ctr_flags |= CTR_FLAG_MAX_RECOVERY_RATE; + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE))) { + if (test_and_set_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) { + rs->ti->error = "Only one max_recovery_rate argument pair allowed"; + return -EINVAL; + } if (value > INT_MAX) { rs->ti->error = "max_recovery_rate out of range"; return -EINVAL; } rs->md.sync_speed_max = (int)value; - } else if (!strcasecmp(key, "region_size")) { - rs->ctr_flags |= CTR_FLAG_REGION_SIZE; + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE))) { + if (test_and_set_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) { + rs->ti->error = "Only one region_size argument pair allowed"; + return -EINVAL; + } + region_size = value; - } else if (!strcasecmp(key, "raid10_copies") && - (rs->raid_type->level == 10)) { - if ((value < 2) || (value > 0xFF)) { + rs->requested_bitmap_chunk_sectors = value; + } else if (!strcasecmp(key, dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES))) { + if (test_and_set_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) { + rs->ti->error = "Only one raid10_copies argument pair allowed"; + return -EINVAL; + } + + if (!__within_range(value, 2, rs->md.raid_disks)) { rs->ti->error = "Bad value for 'raid10_copies'"; return -EINVAL; } - rs->ctr_flags |= CTR_FLAG_RAID10_COPIES; + raid10_copies = value; } else { DMERR("Unable to parse RAID parameter: %s", key); - rs->ti->error = "Unable to parse RAID parameters"; + rs->ti->error = "Unable to parse RAID parameter"; return -EINVAL; } } + if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) && + test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { + rs->ti->error = "sync and nosync are mutually exclusive"; + return -EINVAL; + } + + if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) && + (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) || + test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags))) { + rs->ti->error = "sync/nosync and rebuild are mutually exclusive"; + return -EINVAL; + } + + if (write_mostly >= rs->md.raid_disks) { + rs->ti->error = "Can't set all raid1 devices to write_mostly"; + return -EINVAL; + } + if (validate_region_size(rs, region_size)) return -EINVAL; @@ -697,47 +1335,193 @@ static int parse_raid_params(struct raid_set *rs, char **argv, if (dm_set_target_max_io_len(rs->ti, max_io_len)) return -EINVAL; - if (rs->raid_type->level == 10) { + if (rt_is_raid10(rt)) { if (raid10_copies > rs->md.raid_disks) { rs->ti->error = "Not enough devices to satisfy specification"; return -EINVAL; } - /* - * If the format is not "near", we only support - * two copies at the moment. - */ - if (strcmp("near", raid10_format) && (raid10_copies > 2)) { - rs->ti->error = "Too many copies for given RAID10 format."; + rs->md.new_layout = raid10_format_to_md_layout(rs, raid10_format, raid10_copies); + if (rs->md.new_layout < 0) { + rs->ti->error = "Error getting raid10 format"; + return rs->md.new_layout; + } + + rt = get_raid_type_by_ll(10, rs->md.new_layout); + if (!rt) { + rs->ti->error = "Failed to recognize new raid10 layout"; return -EINVAL; } - /* (Len * #mirrors) / #devices */ - sectors_per_dev = rs->ti->len * raid10_copies; - sector_div(sectors_per_dev, rs->md.raid_disks); - - rs->md.layout = raid10_format_to_md_layout(raid10_format, - raid10_copies); - rs->md.new_layout = rs->md.layout; - } else if ((!rs->raid_type->level || rs->raid_type->level > 1) && - sector_div(sectors_per_dev, - (rs->md.raid_disks - rs->raid_type->parity_devs))) { - rs->ti->error = "Target length not divisible by number of data devices"; - return -EINVAL; + if ((rt->algorithm == ALGORITHM_RAID10_DEFAULT || + rt->algorithm == ALGORITHM_RAID10_NEAR) && + test_bit(__CTR_FLAG_RAID10_USE_NEAR_SETS, &rs->ctr_flags)) { + rs->ti->error = "RAID10 format 'near' and 'raid10_use_near_sets' are incompatible"; + return -EINVAL; + } } - rs->md.dev_sectors = sectors_per_dev; + + rs->raid10_copies = raid10_copies; /* Assume there are no metadata devices until the drives are parsed */ rs->md.persistent = 0; rs->md.external = 1; + /* Check, if any invalid ctr arguments have been passed in for the raid level */ + return rs_check_for_valid_flags(rs); +} + +/* Set raid4/5/6 cache size */ +static int rs_set_raid456_stripe_cache(struct raid_set *rs) +{ + int r; + struct r5conf *conf; + struct mddev *mddev = &rs->md; + uint32_t min_stripes = max(mddev->chunk_sectors, mddev->new_chunk_sectors) / 2; + uint32_t nr_stripes = rs->stripe_cache_entries; + + if (!rt_is_raid456(rs->raid_type)) { + rs->ti->error = "Inappropriate raid level; cannot change stripe_cache size"; + return -EINVAL; + } + + if (nr_stripes < min_stripes) { + DMINFO("Adjusting requested %u stripe cache entries to %u to suit stripe size", + nr_stripes, min_stripes); + nr_stripes = min_stripes; + } + + conf = mddev->private; + if (!conf) { + rs->ti->error = "Cannot change stripe_cache size on inactive RAID set"; + return -EINVAL; + } + + /* Try setting number of stripes in raid456 stripe cache */ + if (conf->min_nr_stripes != nr_stripes) { + r = raid5_set_cache_size(mddev, nr_stripes); + if (r) { + rs->ti->error = "Failed to set raid4/5/6 stripe cache size"; + return r; + } + + DMINFO("%u stripe cache entries", nr_stripes); + } + return 0; } +/* Return # of data stripes as kept in mddev as of @rs (i.e. as of superblock) */ +static unsigned int mddev_data_stripes(struct raid_set *rs) +{ + return rs->md.raid_disks - rs->raid_type->parity_devs; +} + +/* Return # of data stripes of @rs (i.e. as of ctr) */ +static unsigned int rs_data_stripes(struct raid_set *rs) +{ + return rs->raid_disks - rs->raid_type->parity_devs; +} + +/* 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) +{ + int delta_disks; + unsigned int data_stripes; + struct mddev *mddev = &rs->md; + struct md_rdev *rdev; + sector_t array_sectors = rs->ti->len, dev_sectors = rs->ti->len; + + if (use_mddev) { + delta_disks = mddev->delta_disks; + data_stripes = mddev_data_stripes(rs); + } else { + delta_disks = rs->delta_disks; + data_stripes = rs_data_stripes(rs); + } + + /* Special raid1 case w/o delta_disks support (yet) */ + if (rt_is_raid1(rs->raid_type)) + ; + else if (rt_is_raid10(rs->raid_type)) { + if (rs->raid10_copies < 2 || + delta_disks < 0) { + rs->ti->error = "Bogus raid10 data copies or delta disks"; + return -EINVAL; + } + + dev_sectors *= rs->raid10_copies; + if (sector_div(dev_sectors, data_stripes)) + goto bad; + + array_sectors = (data_stripes + delta_disks) * dev_sectors; + if (sector_div(array_sectors, rs->raid10_copies)) + goto bad; + + } else if (sector_div(dev_sectors, data_stripes)) + goto bad; + + else + /* Striped layouts */ + array_sectors = (data_stripes + delta_disks) * dev_sectors; + + rdev_for_each(rdev, mddev) + rdev->sectors = dev_sectors; + + mddev->array_sectors = array_sectors; + mddev->dev_sectors = dev_sectors; + + return 0; +bad: + rs->ti->error = "Target length not divisible by number of data devices"; + return -EINVAL; +} + +/* Setup recovery on @rs */ +static void __rs_setup_recovery(struct raid_set *rs, sector_t dev_sectors) +{ + /* raid0 does not recover */ + if (rs_is_raid0(rs)) + rs->md.recovery_cp = MaxSector; + /* + * A raid6 set has to be recovered either + * completely or for the grown part to + * ensure proper parity and Q-Syndrome + */ + else if (rs_is_raid6(rs)) + rs->md.recovery_cp = dev_sectors; + /* + * Other raid set types may skip recovery + * depending on the 'nosync' flag. + */ + else + rs->md.recovery_cp = test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags) + ? MaxSector : dev_sectors; +} + +/* Setup recovery on @rs based on raid type, device size and 'nosync' flag */ +static void rs_setup_recovery(struct raid_set *rs, sector_t dev_sectors) +{ + if (!dev_sectors) + /* New raid set or 'sync' flag provided */ + __rs_setup_recovery(rs, 0); + else if (dev_sectors == MaxSector) + /* Prevent recovery */ + __rs_setup_recovery(rs, MaxSector); + else if (rs->dev[0].rdev.sectors < dev_sectors) + /* Grown raid set */ + __rs_setup_recovery(rs, rs->dev[0].rdev.sectors); + else + __rs_setup_recovery(rs, MaxSector); +} + 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)) + rs_set_capacity(rs); dm_table_event(rs->ti->table); } @@ -749,19 +1533,225 @@ static int raid_is_congested(struct dm_target_callbacks *cb, int bits) } /* + * Make sure a valid takover (level switch) is being requested on @rs + * + * Conversions of raid sets from one MD personality to another + * have to conform to restrictions which are enforced here. + */ +static int rs_check_takeover(struct raid_set *rs) +{ + struct mddev *mddev = &rs->md; + unsigned int near_copies; + + if (rs->md.degraded) { + rs->ti->error = "Can't takeover degraded raid set"; + return -EPERM; + } + + if (rs_is_reshaping(rs)) { + rs->ti->error = "Can't takeover reshaping raid set"; + return -EPERM; + } + + switch (mddev->level) { + case 0: + /* raid0 -> raid1/5 with one disk */ + if ((mddev->new_level == 1 || mddev->new_level == 5) && + mddev->raid_disks == 1) + return 0; + + /* raid0 -> raid10 */ + if (mddev->new_level == 10 && + !(rs->raid_disks % mddev->raid_disks)) + return 0; + + /* raid0 with multiple disks -> raid4/5/6 */ + if (__within_range(mddev->new_level, 4, 6) && + mddev->new_layout == ALGORITHM_PARITY_N && + mddev->raid_disks > 1) + return 0; + + break; + + case 10: + /* Can't takeover raid10_offset! */ + if (__is_raid10_offset(mddev->layout)) + break; + + near_copies = __raid10_near_copies(mddev->layout); + + /* raid10* -> raid0 */ + if (mddev->new_level == 0) { + /* Can takeover raid10_near with raid disks divisable by data copies! */ + if (near_copies > 1 && + !(mddev->raid_disks % near_copies)) { + mddev->raid_disks /= near_copies; + mddev->delta_disks = mddev->raid_disks; + return 0; + } + + /* Can takeover raid10_far */ + if (near_copies == 1 && + __raid10_far_copies(mddev->layout) > 1) + return 0; + + break; + } + + /* raid10_{near,far} -> raid1 */ + if (mddev->new_level == 1 && + max(near_copies, __raid10_far_copies(mddev->layout)) == mddev->raid_disks) + return 0; + + /* raid10_{near,far} with 2 disks -> raid4/5 */ + if (__within_range(mddev->new_level, 4, 5) && + mddev->raid_disks == 2) + return 0; + break; + + case 1: + /* raid1 with 2 disks -> raid4/5 */ + if (__within_range(mddev->new_level, 4, 5) && + mddev->raid_disks == 2) { + mddev->degraded = 1; + return 0; + } + + /* raid1 -> raid0 */ + if (mddev->new_level == 0 && + mddev->raid_disks == 1) + return 0; + + /* raid1 -> raid10 */ + if (mddev->new_level == 10) + return 0; + break; + + case 4: + /* raid4 -> raid0 */ + if (mddev->new_level == 0) + return 0; + + /* raid4 -> raid1/5 with 2 disks */ + if ((mddev->new_level == 1 || mddev->new_level == 5) && + mddev->raid_disks == 2) + return 0; + + /* raid4 -> raid5/6 with parity N */ + if (__within_range(mddev->new_level, 5, 6) && + mddev->layout == ALGORITHM_PARITY_N) + return 0; + break; + + case 5: + /* raid5 with parity N -> raid0 */ + if (mddev->new_level == 0 && + mddev->layout == ALGORITHM_PARITY_N) + return 0; + + /* raid5 with parity N -> raid4 */ + if (mddev->new_level == 4 && + mddev->layout == ALGORITHM_PARITY_N) + return 0; + + /* raid5 with 2 disks -> raid1/4/10 */ + if ((mddev->new_level == 1 || mddev->new_level == 4 || mddev->new_level == 10) && + mddev->raid_disks == 2) + return 0; + + /* raid5_* -> raid6_*_6 with Q-Syndrome N (e.g. raid5_ra -> raid6_ra_6 */ + if (mddev->new_level == 6 && + ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || + __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC_6, ALGORITHM_RIGHT_SYMMETRIC_6))) + return 0; + break; + + case 6: + /* raid6 with parity N -> raid0 */ + if (mddev->new_level == 0 && + mddev->layout == ALGORITHM_PARITY_N) + return 0; + + /* raid6 with parity N -> raid4 */ + if (mddev->new_level == 4 && + mddev->layout == ALGORITHM_PARITY_N) + return 0; + + /* raid6_*_n with Q-Syndrome N -> raid5_* */ + if (mddev->new_level == 5 && + ((mddev->layout == ALGORITHM_PARITY_N && mddev->new_layout == ALGORITHM_PARITY_N) || + __within_range(mddev->new_layout, ALGORITHM_LEFT_ASYMMETRIC, ALGORITHM_RIGHT_SYMMETRIC))) + return 0; + + default: + break; + } + + rs->ti->error = "takeover not possible"; + return -EINVAL; +} + +/* True if @rs requested to be taken over */ +static bool rs_takeover_requested(struct raid_set *rs) +{ + return rs->md.new_level != rs->md.level; +} + +/* True if @rs is requested to reshape by ctr */ +static bool rs_reshape_requested(struct raid_set *rs) +{ + bool change; + struct mddev *mddev = &rs->md; + + if (rs_takeover_requested(rs)) + return false; + + if (!mddev->level) + return false; + + change = mddev->new_layout != mddev->layout || + mddev->new_chunk_sectors != mddev->chunk_sectors || + rs->delta_disks; + + /* Historical case to support raid1 reshape without delta disks */ + if (mddev->level == 1) { + if (rs->delta_disks) + return !!rs->delta_disks; + + return !change && + mddev->raid_disks != rs->raid_disks; + } + + if (mddev->level == 10) + return change && + !__is_raid10_far(mddev->new_layout) && + rs->delta_disks >= 0; + + return change; +} + +/* Features */ +#define FEATURE_FLAG_SUPPORTS_V190 0x1 /* Supports extended superblock */ + +/* State flags for sb->flags */ +#define SB_FLAG_RESHAPE_ACTIVE 0x1 +#define SB_FLAG_RESHAPE_BACKWARDS 0x2 + +/* * This structure is never routinely used by userspace, unlike md superblocks. * Devices with this superblock should only ever be accessed via device-mapper. */ #define DM_RAID_MAGIC 0x64526D44 struct dm_raid_superblock { __le32 magic; /* "DmRd" */ - __le32 features; /* Used to indicate possible future changes */ + __le32 compat_features; /* Used to indicate compatible features (like 1.9.0 ondisk metadata extension) */ - __le32 num_devices; /* Number of devices in this array. (Max 64) */ - __le32 array_position; /* The position of this drive in the array */ + __le32 num_devices; /* Number of devices in this raid set. (Max 64) */ + __le32 array_position; /* The position of this drive in the raid set */ __le64 events; /* Incremented by md when superblock updated */ - __le64 failed_devices; /* Bit field of devices to indicate failures */ + __le64 failed_devices; /* Pre 1.9.0 part of bit field of devices to */ + /* indicate failures (see extension below) */ /* * This offset tracks the progress of the repair or replacement of @@ -770,21 +1760,95 @@ struct dm_raid_superblock { __le64 disk_recovery_offset; /* - * This offset tracks the progress of the initial array + * This offset tracks the progress of the initial raid set * synchronisation/parity calculation. */ __le64 array_resync_offset; /* - * RAID characteristics + * raid characteristics */ __le32 level; __le32 layout; __le32 stripe_sectors; - /* Remainder of a logical block is zero-filled when writing (see super_sync()). */ + /******************************************************************** + * BELOW FOLLOW V1.9.0 EXTENSIONS TO THE PRISTINE SUPERBLOCK FORMAT!!! + * + * FEATURE_FLAG_SUPPORTS_V190 in the features member indicates that those exist + */ + + __le32 flags; /* Flags defining array states for reshaping */ + + /* + * This offset tracks the progress of a raid + * set reshape in order to be able to restart it + */ + __le64 reshape_position; + + /* + * These define the properties of the array in case of an interrupted reshape + */ + __le32 new_level; + __le32 new_layout; + __le32 new_stripe_sectors; + __le32 delta_disks; + + __le64 array_sectors; /* Array size in sectors */ + + /* + * Sector offsets to data on devices (reshaping). + * Needed to support out of place reshaping, thus + * not writing over any stripes whilst converting + * them from old to new layout + */ + __le64 data_offset; + __le64 new_data_offset; + + __le64 sectors; /* Used device size in sectors */ + + /* + * Additonal Bit field of devices indicating failures to support + * up to 256 devices with the 1.9.0 on-disk metadata format + */ + __le64 extended_failed_devices[DISKS_ARRAY_ELEMS - 1]; + + __le32 incompat_features; /* Used to indicate any incompatible features */ + + /* Always set rest up to logical block size to 0 when writing (see get_metadata_device() below). */ } __packed; +/* + * Check for reshape constraints on raid set @rs: + * + * - reshape function non-existent + * - degraded set + * - ongoing recovery + * - ongoing reshape + * + * Returns 0 if none or -EPERM if given constraint + * and error message reference in @errmsg + */ +static int rs_check_reshape(struct raid_set *rs) +{ + struct mddev *mddev = &rs->md; + + if (!mddev->pers || !mddev->pers->check_reshape) + rs->ti->error = "Reshape not supported"; + else if (mddev->degraded) + rs->ti->error = "Can't reshape degraded raid set"; + else if (rs_is_recovering(rs)) + rs->ti->error = "Convert request on recovering raid set prohibited"; + else if (rs_is_reshaping(rs)) + rs->ti->error = "raid set already reshaping!"; + else if (!(rs_is_raid1(rs) || rs_is_raid10(rs) || rs_is_raid456(rs))) + rs->ti->error = "Reshaping only supported for raid1/4/5/6/10"; + else + return 0; + + return -EPERM; +} + static int read_disk_sb(struct md_rdev *rdev, int size) { BUG_ON(!rdev->sb_page); @@ -792,7 +1856,7 @@ static int read_disk_sb(struct md_rdev *rdev, int size) if (rdev->sb_loaded) return 0; - if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, 0, 1)) { + if (!sync_page_io(rdev, 0, size, rdev->sb_page, REQ_OP_READ, 0, true)) { DMERR("Failed to read superblock of device at position %d", rdev->raid_disk); md_error(rdev->mddev, rdev); @@ -804,31 +1868,67 @@ static int read_disk_sb(struct md_rdev *rdev, int size) return 0; } +static void sb_retrieve_failed_devices(struct dm_raid_superblock *sb, uint64_t *failed_devices) +{ + failed_devices[0] = le64_to_cpu(sb->failed_devices); + memset(failed_devices + 1, 0, sizeof(sb->extended_failed_devices)); + + if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { + int i = ARRAY_SIZE(sb->extended_failed_devices); + + while (i--) + failed_devices[i+1] = le64_to_cpu(sb->extended_failed_devices[i]); + } +} + +static void sb_update_failed_devices(struct dm_raid_superblock *sb, uint64_t *failed_devices) +{ + int i = ARRAY_SIZE(sb->extended_failed_devices); + + sb->failed_devices = cpu_to_le64(failed_devices[0]); + while (i--) + sb->extended_failed_devices[i] = cpu_to_le64(failed_devices[i+1]); +} + +/* + * Synchronize the superblock members with the raid set properties + * + * All superblock data is little endian. + */ static void super_sync(struct mddev *mddev, struct md_rdev *rdev) { - int i; - uint64_t failed_devices; + bool update_failed_devices = false; + unsigned int i; + uint64_t failed_devices[DISKS_ARRAY_ELEMS]; struct dm_raid_superblock *sb; struct raid_set *rs = container_of(mddev, struct raid_set, md); + /* No metadata device, no superblock */ + if (!rdev->meta_bdev) + return; + + BUG_ON(!rdev->sb_page); + sb = page_address(rdev->sb_page); - failed_devices = le64_to_cpu(sb->failed_devices); - for (i = 0; i < mddev->raid_disks; i++) - if (!rs->dev[i].data_dev || - test_bit(Faulty, &(rs->dev[i].rdev.flags))) - failed_devices |= (1ULL << i); + sb_retrieve_failed_devices(sb, failed_devices); - memset(sb + 1, 0, rdev->sb_size - sizeof(*sb)); + for (i = 0; i < rs->raid_disks; i++) + if (!rs->dev[i].data_dev || test_bit(Faulty, &rs->dev[i].rdev.flags)) { + update_failed_devices = true; + set_bit(i, (void *) failed_devices); + } + + if (update_failed_devices) + sb_update_failed_devices(sb, failed_devices); sb->magic = cpu_to_le32(DM_RAID_MAGIC); - sb->features = cpu_to_le32(0); /* No features yet */ + sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); sb->num_devices = cpu_to_le32(mddev->raid_disks); sb->array_position = cpu_to_le32(rdev->raid_disk); sb->events = cpu_to_le64(mddev->events); - sb->failed_devices = cpu_to_le64(failed_devices); sb->disk_recovery_offset = cpu_to_le64(rdev->recovery_offset); sb->array_resync_offset = cpu_to_le64(mddev->recovery_cp); @@ -836,6 +1936,33 @@ static void super_sync(struct mddev *mddev, struct md_rdev *rdev) sb->level = cpu_to_le32(mddev->level); sb->layout = cpu_to_le32(mddev->layout); sb->stripe_sectors = cpu_to_le32(mddev->chunk_sectors); + + sb->new_level = cpu_to_le32(mddev->new_level); + sb->new_layout = cpu_to_le32(mddev->new_layout); + sb->new_stripe_sectors = cpu_to_le32(mddev->new_chunk_sectors); + + sb->delta_disks = cpu_to_le32(mddev->delta_disks); + + smp_rmb(); /* Make sure we access most recent reshape position */ + sb->reshape_position = cpu_to_le64(mddev->reshape_position); + if (le64_to_cpu(sb->reshape_position) != MaxSector) { + /* Flag ongoing reshape */ + sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE); + + if (mddev->delta_disks < 0 || mddev->reshape_backwards) + sb->flags |= cpu_to_le32(SB_FLAG_RESHAPE_BACKWARDS); + } else { + /* Clear reshape flags */ + sb->flags &= ~(cpu_to_le32(SB_FLAG_RESHAPE_ACTIVE|SB_FLAG_RESHAPE_BACKWARDS)); + } + + sb->array_sectors = cpu_to_le64(mddev->array_sectors); + sb->data_offset = cpu_to_le64(rdev->data_offset); + sb->new_data_offset = cpu_to_le64(rdev->new_data_offset); + sb->sectors = cpu_to_le64(rdev->sectors); + + /* Zero out the rest of the payload after the size of the superblock */ + memset(sb + 1, 0, rdev->sb_size - sizeof(*sb)); } /* @@ -848,7 +1975,7 @@ static void super_sync(struct mddev *mddev, struct md_rdev *rdev) */ static int super_load(struct md_rdev *rdev, struct md_rdev *refdev) { - int ret; + int r; struct dm_raid_superblock *sb; struct dm_raid_superblock *refsb; uint64_t events_sb, events_refsb; @@ -860,9 +1987,9 @@ static int super_load(struct md_rdev *rdev, struct md_rdev *refdev) return -EINVAL; } - ret = read_disk_sb(rdev, rdev->sb_size); - if (ret) - return ret; + r = read_disk_sb(rdev, rdev->sb_size); + if (r) + return r; sb = page_address(rdev->sb_page); @@ -876,6 +2003,7 @@ static int super_load(struct md_rdev *rdev, struct md_rdev *refdev) super_sync(rdev->mddev, rdev); set_bit(FirstUse, &rdev->flags); + sb->compat_features = cpu_to_le32(FEATURE_FLAG_SUPPORTS_V190); /* Force writing of superblocks to disk */ set_bit(MD_CHANGE_DEVS, &rdev->mddev->flags); @@ -895,129 +2023,212 @@ static int super_load(struct md_rdev *rdev, struct md_rdev *refdev) return (events_sb > events_refsb) ? 1 : 0; } -static int super_init_validation(struct mddev *mddev, struct md_rdev *rdev) +static int super_init_validation(struct raid_set *rs, struct md_rdev *rdev) { int role; - struct raid_set *rs = container_of(mddev, struct raid_set, md); + unsigned int d; + struct mddev *mddev = &rs->md; uint64_t events_sb; - uint64_t failed_devices; + uint64_t failed_devices[DISKS_ARRAY_ELEMS]; struct dm_raid_superblock *sb; - uint32_t new_devs = 0; - uint32_t rebuilds = 0; + uint32_t new_devs = 0, rebuild_and_new = 0, rebuilds = 0; struct md_rdev *r; struct dm_raid_superblock *sb2; sb = page_address(rdev->sb_page); events_sb = le64_to_cpu(sb->events); - failed_devices = le64_to_cpu(sb->failed_devices); /* * Initialise to 1 if this is a new superblock. */ mddev->events = events_sb ? : 1; + mddev->reshape_position = MaxSector; + /* - * Reshaping is not currently allowed + * Reshaping is supported, e.g. reshape_position is valid + * in superblock and superblock content is authoritative. */ - if (le32_to_cpu(sb->level) != mddev->level) { - DMERR("Reshaping arrays not yet supported. (RAID level change)"); - return -EINVAL; - } - if (le32_to_cpu(sb->layout) != mddev->layout) { - DMERR("Reshaping arrays not yet supported. (RAID layout change)"); - DMERR(" 0x%X vs 0x%X", le32_to_cpu(sb->layout), mddev->layout); - DMERR(" Old layout: %s w/ %d copies", - raid10_md_layout_to_format(le32_to_cpu(sb->layout)), - raid10_md_layout_to_copies(le32_to_cpu(sb->layout))); - DMERR(" New layout: %s w/ %d copies", - raid10_md_layout_to_format(mddev->layout), - raid10_md_layout_to_copies(mddev->layout)); - return -EINVAL; - } - if (le32_to_cpu(sb->stripe_sectors) != mddev->chunk_sectors) { - DMERR("Reshaping arrays not yet supported. (stripe sectors change)"); - return -EINVAL; - } + if (le32_to_cpu(sb->compat_features) & FEATURE_FLAG_SUPPORTS_V190) { + /* Superblock is authoritative wrt given raid set layout! */ + mddev->raid_disks = le32_to_cpu(sb->num_devices); + mddev->level = le32_to_cpu(sb->level); + mddev->layout = le32_to_cpu(sb->layout); + mddev->chunk_sectors = le32_to_cpu(sb->stripe_sectors); + mddev->new_level = le32_to_cpu(sb->new_level); + mddev->new_layout = le32_to_cpu(sb->new_layout); + mddev->new_chunk_sectors = le32_to_cpu(sb->new_stripe_sectors); + mddev->delta_disks = le32_to_cpu(sb->delta_disks); + mddev->array_sectors = le64_to_cpu(sb->array_sectors); + + /* raid was reshaping and got interrupted */ + if (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_ACTIVE) { + if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) { + DMERR("Reshape requested but raid set is still reshaping"); + return -EINVAL; + } - /* We can only change the number of devices in RAID1 right now */ - if ((rs->raid_type->level != 1) && - (le32_to_cpu(sb->num_devices) != mddev->raid_disks)) { - DMERR("Reshaping arrays not yet supported. (device count change)"); - return -EINVAL; + if (mddev->delta_disks < 0 || + (!mddev->delta_disks && (le32_to_cpu(sb->flags) & SB_FLAG_RESHAPE_BACKWARDS))) + mddev->reshape_backwards = 1; + else + mddev->reshape_backwards = 0; + + mddev->reshape_position = le64_to_cpu(sb->reshape_position); + rs->raid_type = get_raid_type_by_ll(mddev->level, mddev->layout); + } + + } else { + /* + * No takeover/reshaping, because we don't have the extended v1.9.0 metadata + */ + if (le32_to_cpu(sb->level) != mddev->level) { + DMERR("Reshaping/takeover raid sets not yet supported. (raid level/stripes/size change)"); + return -EINVAL; + } + if (le32_to_cpu(sb->layout) != mddev->layout) { + DMERR("Reshaping raid sets not yet supported. (raid layout change)"); + DMERR(" 0x%X vs 0x%X", le32_to_cpu(sb->layout), mddev->layout); + DMERR(" Old layout: %s w/ %d copies", + raid10_md_layout_to_format(le32_to_cpu(sb->layout)), + raid10_md_layout_to_copies(le32_to_cpu(sb->layout))); + DMERR(" New layout: %s w/ %d copies", + raid10_md_layout_to_format(mddev->layout), + raid10_md_layout_to_copies(mddev->layout)); + return -EINVAL; + } + if (le32_to_cpu(sb->stripe_sectors) != mddev->chunk_sectors) { + DMERR("Reshaping raid sets not yet supported. (stripe sectors change)"); + return -EINVAL; + } + + /* We can only change the number of devices in raid1 with old (i.e. pre 1.0.7) metadata */ + if (!rt_is_raid1(rs->raid_type) && + (le32_to_cpu(sb->num_devices) != mddev->raid_disks)) { + DMERR("Reshaping raid sets not yet supported. (device count change from %u to %u)", + sb->num_devices, mddev->raid_disks); + return -EINVAL; + } + + /* Table line is checked vs. authoritative superblock */ + rs_set_new(rs); } - if (!(rs->ctr_flags & (CTR_FLAG_SYNC | CTR_FLAG_NOSYNC))) + if (!test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) mddev->recovery_cp = le64_to_cpu(sb->array_resync_offset); /* * During load, we set FirstUse if a new superblock was written. * There are two reasons we might not have a superblock: - * 1) The array is brand new - in which case, all of the - * devices must have their In_sync bit set. Also, + * 1) The raid set is brand new - in which case, all of the + * devices must have their In_sync bit set. Also, * recovery_cp must be 0, unless forced. - * 2) This is a new device being added to an old array + * 2) This is a new device being added to an old raid set * and the new device needs to be rebuilt - in which * case the In_sync bit will /not/ be set and * recovery_cp must be MaxSector. + * 3) This is/are a new device(s) being added to an old + * raid set during takeover to a higher raid level + * to provide capacity for redundancy or during reshape + * to add capacity to grow the raid set. */ + d = 0; rdev_for_each(r, mddev) { + if (test_bit(FirstUse, &r->flags)) + new_devs++; + if (!test_bit(In_sync, &r->flags)) { - DMINFO("Device %d specified for rebuild: " - "Clearing superblock", r->raid_disk); + DMINFO("Device %d specified for rebuild; clearing superblock", + r->raid_disk); rebuilds++; - } else if (test_bit(FirstUse, &r->flags)) - new_devs++; + + if (test_bit(FirstUse, &r->flags)) + rebuild_and_new++; + } + + d++; } - if (!rebuilds) { - if (new_devs == mddev->raid_disks) { - DMINFO("Superblocks created for new array"); + if (new_devs == rs->raid_disks || !rebuilds) { + /* Replace a broken device */ + if (new_devs == 1 && !rs->delta_disks) + ; + if (new_devs == rs->raid_disks) { + DMINFO("Superblocks created for new raid set"); set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); - } else if (new_devs) { - DMERR("New device injected " - "into existing array without 'rebuild' " - "parameter specified"); + } else if (new_devs != rebuilds && + new_devs != rs->delta_disks) { + DMERR("New device injected into existing raid set without " + "'delta_disks' or 'rebuild' parameter specified"); return -EINVAL; } - } else if (new_devs) { - DMERR("'rebuild' devices cannot be " - "injected into an array with other first-time devices"); - return -EINVAL; - } else if (mddev->recovery_cp != MaxSector) { - DMERR("'rebuild' specified while array is not in-sync"); + } else if (new_devs && new_devs != rebuilds) { + DMERR("%u 'rebuild' devices cannot be injected into" + " a raid set with %u other first-time devices", + rebuilds, new_devs); return -EINVAL; + } else if (rebuilds) { + if (rebuild_and_new && rebuilds != rebuild_and_new) { + DMERR("new device%s provided without 'rebuild'", + new_devs > 1 ? "s" : ""); + return -EINVAL; + } else if (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; + } else if (rs_is_reshaping(rs)) { + DMERR("'rebuild' specified while raid set is being reshaped (reshape_position=%llu)", + (unsigned long long) mddev->reshape_position); + return -EINVAL; + } } /* * Now we set the Faulty bit for those devices that are * recorded in the superblock as failed. */ + sb_retrieve_failed_devices(sb, failed_devices); rdev_for_each(r, mddev) { if (!r->sb_page) continue; sb2 = page_address(r->sb_page); sb2->failed_devices = 0; + memset(sb2->extended_failed_devices, 0, sizeof(sb2->extended_failed_devices)); /* * Check for any device re-ordering. */ if (!test_bit(FirstUse, &r->flags) && (r->raid_disk >= 0)) { role = le32_to_cpu(sb2->array_position); + if (role < 0) + continue; + if (role != r->raid_disk) { - if (rs->raid_type->level != 1) { - rs->ti->error = "Cannot change device " - "positions in RAID array"; + if (__is_raid10_near(mddev->layout)) { + if (mddev->raid_disks % __raid10_near_copies(mddev->layout) || + rs->raid_disks % rs->raid10_copies) { + rs->ti->error = + "Cannot change raid10 near set to odd # of devices!"; + return -EINVAL; + } + + sb2->array_position = cpu_to_le32(r->raid_disk); + + } else if (!(rs_is_raid10(rs) && rt_is_raid0(rs->raid_type)) && + !(rs_is_raid0(rs) && rt_is_raid10(rs->raid_type)) && + !rt_is_raid1(rs->raid_type)) { + rs->ti->error = "Cannot change device positions in raid set"; return -EINVAL; } - DMINFO("RAID1 device #%d now at position #%d", - role, r->raid_disk); + + DMINFO("raid device #%d now at position #%d", role, r->raid_disk); } /* * Partial recovery is performed on * returning failed devices. */ - if (failed_devices & (1 << role)) + if (test_bit(role, (void *) failed_devices)) set_bit(Faulty, &r->flags); } } @@ -1028,41 +2239,60 @@ static int super_init_validation(struct mddev *mddev, struct md_rdev *rdev) static int super_validate(struct raid_set *rs, struct md_rdev *rdev) { struct mddev *mddev = &rs->md; - struct dm_raid_superblock *sb = page_address(rdev->sb_page); + struct dm_raid_superblock *sb; + + if (rs_is_raid0(rs) || !rdev->sb_page) + return 0; + + sb = page_address(rdev->sb_page); /* * If mddev->events is not set, we know we have not yet initialized * the array. */ - if (!mddev->events && super_init_validation(mddev, rdev)) + if (!mddev->events && super_init_validation(rs, rdev)) return -EINVAL; - if (le32_to_cpu(sb->features)) { - rs->ti->error = "Unable to assemble array: No feature flags supported yet"; + if (le32_to_cpu(sb->compat_features) != FEATURE_FLAG_SUPPORTS_V190) { + rs->ti->error = "Unable to assemble array: Unknown flag(s) in compatible feature flags"; + return -EINVAL; + } + + if (sb->incompat_features) { + rs->ti->error = "Unable to assemble array: No incompatible feature flags supported yet"; return -EINVAL; } /* Enable bitmap creation for RAID levels != 0 */ - mddev->bitmap_info.offset = (rs->raid_type->level) ? to_sector(4096) : 0; + mddev->bitmap_info.offset = rt_is_raid0(rs->raid_type) ? 0 : to_sector(4096); rdev->mddev->bitmap_info.default_offset = mddev->bitmap_info.offset; - if (!test_bit(FirstUse, &rdev->flags)) { + if (!test_and_clear_bit(FirstUse, &rdev->flags)) { + /* Retrieve device size stored in superblock to be prepared for shrink */ + rdev->sectors = le64_to_cpu(sb->sectors); rdev->recovery_offset = le64_to_cpu(sb->disk_recovery_offset); - if (rdev->recovery_offset != MaxSector) - clear_bit(In_sync, &rdev->flags); + if (rdev->recovery_offset == MaxSector) + set_bit(In_sync, &rdev->flags); + /* + * If no reshape in progress -> we're recovering single + * disk(s) and have to set the device(s) to out-of-sync + */ + else if (!rs_is_reshaping(rs)) + clear_bit(In_sync, &rdev->flags); /* Mandatory for recovery */ } /* * If a device comes back, set it as not In_sync and no longer faulty. */ - if (test_bit(Faulty, &rdev->flags)) { - clear_bit(Faulty, &rdev->flags); + if (test_and_clear_bit(Faulty, &rdev->flags)) { + rdev->recovery_offset = 0; clear_bit(In_sync, &rdev->flags); rdev->saved_raid_disk = rdev->raid_disk; - rdev->recovery_offset = 0; } - clear_bit(FirstUse, &rdev->flags); + /* Reshape support -> restore repective data offsets */ + rdev->data_offset = le64_to_cpu(sb->data_offset); + rdev->new_data_offset = le64_to_cpu(sb->new_data_offset); return 0; } @@ -1072,7 +2302,7 @@ static int super_validate(struct raid_set *rs, struct md_rdev *rdev) */ static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs) { - int ret; + int r; struct raid_dev *dev; struct md_rdev *rdev, *tmp, *freshest; struct mddev *mddev = &rs->md; @@ -1082,24 +2312,22 @@ static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs) /* * Skipping super_load due to CTR_FLAG_SYNC will cause * the array to undergo initialization again as - * though it were new. This is the intended effect + * though it were new. This is the intended effect * of the "sync" directive. * * When reshaping capability is added, we must ensure * that the "sync" directive is disallowed during the * reshape. */ - rdev->sectors = to_sector(i_size_read(rdev->bdev->bd_inode)); - - if (rs->ctr_flags & CTR_FLAG_SYNC) + if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) continue; if (!rdev->meta_bdev) continue; - ret = super_load(rdev, freshest); + r = super_load(rdev, freshest); - switch (ret) { + switch (r) { case 1: freshest = rdev; break; @@ -1148,25 +2376,336 @@ static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs) * Validation of the freshest device provides the source of * validation for the remaining devices. */ - ti->error = "Unable to assemble array: Invalid superblocks"; + rs->ti->error = "Unable to assemble array: Invalid superblocks"; if (super_validate(rs, freshest)) return -EINVAL; rdev_for_each(rdev, mddev) if ((rdev != freshest) && super_validate(rs, rdev)) return -EINVAL; + return 0; +} + +/* + * Adjust data_offset and new_data_offset on all disk members of @rs + * for out of place reshaping if requested by contructor + * + * We need free space at the beginning of each raid disk for forward + * and at the end for backward reshapes which userspace has to provide + * via remapping/reordering of space. + */ +static int rs_adjust_data_offsets(struct raid_set *rs) +{ + sector_t data_offset = 0, new_data_offset = 0; + struct md_rdev *rdev; + + /* Constructor did not request data offset change */ + if (!test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) { + if (!rs_is_reshapable(rs)) + goto out; + + return 0; + } + + /* HM FIXME: get InSync raid_dev? */ + rdev = &rs->dev[0].rdev; + + if (rs->delta_disks < 0) { + /* + * Removing disks (reshaping backwards): + * + * - before reshape: data is at offset 0 and free space + * is at end of each component LV + * + * - after reshape: data is at offset rs->data_offset != 0 on each component LV + */ + data_offset = 0; + new_data_offset = rs->data_offset; + + } else if (rs->delta_disks > 0) { + /* + * Adding disks (reshaping forwards): + * + * - before reshape: data is at offset rs->data_offset != 0 and + * free space is at begin of each component LV + * + * - after reshape: data is at offset 0 on each component LV + */ + data_offset = rs->data_offset; + new_data_offset = 0; + + } else { + /* + * User space passes in 0 for data offset after having removed reshape space + * + * - or - (data offset != 0) + * + * Changing RAID layout or chunk size -> toggle offsets + * + * - before reshape: data is at offset rs->data_offset 0 and + * free space is at end of each component LV + * -or- + * data is at offset rs->data_offset != 0 and + * free space is at begin of each component LV + * + * - after reshape: data is at offset 0 if it was at offset != 0 + * or at offset != 0 if it was at offset 0 + * on each component LV + * + */ + data_offset = rs->data_offset ? rdev->data_offset : 0; + new_data_offset = data_offset ? 0 : rs->data_offset; + set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); + } + + /* + * 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) { + rs->ti->error = data_offset ? "No space for forward reshape" : + "No space for backward reshape"; + return -ENOSPC; + } +out: + /* Adjust data offsets on all rdevs */ + rdev_for_each(rdev, &rs->md) { + rdev->data_offset = data_offset; + rdev->new_data_offset = new_data_offset; + } + + return 0; +} + +/* Userpace reordered disks -> adjust raid_disk indexes in @rs */ +static void __reorder_raid_disk_indexes(struct raid_set *rs) +{ + int i = 0; + struct md_rdev *rdev; + + rdev_for_each(rdev, &rs->md) { + rdev->raid_disk = i++; + rdev->saved_raid_disk = rdev->new_raid_disk = -1; + } +} + +/* + * Setup @rs for takeover by a different raid level + */ +static int rs_setup_takeover(struct raid_set *rs) +{ + struct mddev *mddev = &rs->md; + struct md_rdev *rdev; + unsigned int d = mddev->raid_disks = rs->raid_disks; + 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) { + /* 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) + /* raid1 -> raid10_near layout */ + mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, + rs->raid_disks); + else + return -EINVAL; + + } + + clear_bit(MD_ARRAY_FIRST_USE, &mddev->flags); + mddev->recovery_cp = MaxSector; + + while (d--) { + rdev = &rs->dev[d].rdev; + + if (test_bit(d, (void *) rs->rebuild_disks)) { + clear_bit(In_sync, &rdev->flags); + clear_bit(Faulty, &rdev->flags); + mddev->recovery_cp = rdev->recovery_offset = 0; + /* Bitmap has to be created when we do an "up" takeover */ + set_bit(MD_ARRAY_FIRST_USE, &mddev->flags); + } + + rdev->new_data_offset = new_data_offset; + } + + return 0; +} + +/* Prepare @rs for reshape */ +static int rs_prepare_reshape(struct raid_set *rs) +{ + bool reshape; + struct mddev *mddev = &rs->md; + + if (rs_is_raid10(rs)) { + if (rs->raid_disks != mddev->raid_disks && + __is_raid10_near(mddev->layout) && + rs->raid10_copies && + rs->raid10_copies != __raid10_near_copies(mddev->layout)) { + /* + * raid disk have to be multiple of data copies to allow this conversion, + * + * This is actually not a reshape it is a + * rebuild of any additional mirrors per group + */ + if (rs->raid_disks % rs->raid10_copies) { + rs->ti->error = "Can't reshape raid10 mirror groups"; + return -EINVAL; + } + + /* Userpace reordered disks to add/remove mirrors -> adjust raid_disk indexes */ + __reorder_raid_disk_indexes(rs); + mddev->layout = raid10_format_to_md_layout(rs, ALGORITHM_RAID10_NEAR, + rs->raid10_copies); + mddev->new_layout = mddev->layout; + reshape = false; + } else + reshape = true; + + } else if (rs_is_raid456(rs)) + reshape = true; + + else if (rs_is_raid1(rs)) { + if (rs->delta_disks) { + /* Process raid1 via delta_disks */ + mddev->degraded = rs->delta_disks < 0 ? -rs->delta_disks : rs->delta_disks; + reshape = true; + } else { + /* Process raid1 without delta_disks */ + mddev->raid_disks = rs->raid_disks; + set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags); + reshape = false; + } + } else { + rs->ti->error = "Called with bogus raid type"; + return -EINVAL; + } + + if (reshape) { + set_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags); + set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); + set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags); + } else if (mddev->raid_disks < rs->raid_disks) + /* Create new superblocks and bitmaps, if any new disks */ + set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); return 0; } /* + * + * - change raid layout + * - change chunk size + * - add disks + * - remove disks + */ +static int rs_setup_reshape(struct raid_set *rs) +{ + int r = 0; + unsigned int cur_raid_devs, d; + struct mddev *mddev = &rs->md; + struct md_rdev *rdev; + + mddev->delta_disks = rs->delta_disks; + cur_raid_devs = mddev->raid_disks; + + /* Ignore impossible layout change whilst adding/removing disks */ + if (mddev->delta_disks && + mddev->layout != mddev->new_layout) { + DMINFO("Ignoring invalid layout change with delta_disks=%d", rs->delta_disks); + mddev->new_layout = mddev->layout; + } + + /* + * Adjust array size: + * + * - in case of adding disks, 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 + * has to shrink before starting the reshape, + * which'll happen here; + * reshape will happen backward, so space has to + * be available at the end of each disk + * + * - data_offset and new_data_offset are + * adjusted for aforementioned out of place + * reshaping based on userspace passing in + * the "data_offset <sectors>" key/value + * pair via the constructor + */ + + /* Add disk(s) */ + if (rs->delta_disks > 0) { + /* Prepare disks for check in raid4/5/6/10 {check|start}_reshape */ + for (d = cur_raid_devs; d < rs->raid_disks; d++) { + rdev = &rs->dev[d].rdev; + clear_bit(In_sync, &rdev->flags); + + /* + * save_raid_disk needs to be -1, or recovery_offset will be set to 0 + * by md, which'll store that erroneously in the superblock on reshape + */ + rdev->saved_raid_disk = -1; + rdev->raid_disk = d; + + rdev->sectors = mddev->dev_sectors; + rdev->recovery_offset = rs_is_raid1(rs) ? 0 : MaxSector; + } + + mddev->reshape_backwards = 0; /* adding disks -> forward reshape */ + + /* Remove disk(s) */ + } else if (rs->delta_disks < 0) { + r = rs_set_dev_and_array_sectors(rs, true); + mddev->reshape_backwards = 1; /* removing disk(s) -> backward reshape */ + + /* Change layout and/or chunk size */ + } else { + /* + * Reshape layout (e.g. raid5_ls -> raid5_n) and/or chunk size: + * + * keeping number of disks and do layout change -> + * + * toggle reshape_backward depending on data_offset: + * + * - free space upfront -> reshape forward + * + * - free space at the end -> reshape backward + * + * + * This utilizes free reshape space avoiding the need + * for userspace to move (parts of) LV segments in + * case of layout/chunksize change (for disk + * adding/removing reshape space has to be at + * the proper address (see above with delta_disks): + * + * add disk(s) -> begin + * remove disk(s)-> end + */ + mddev->reshape_backwards = rs->dev[0].rdev.data_offset ? 0 : 1; + } + + return r; +} + +/* * Enable/disable discard support on RAID set depending on * RAID level and discard properties of underlying RAID members. */ -static void configure_discard_support(struct dm_target *ti, struct raid_set *rs) +static void configure_discard_support(struct raid_set *rs) { int i; bool raid456; + struct dm_target *ti = rs->ti; /* Assume discards not supported until after checks below. */ ti->discards_supported = false; @@ -1174,7 +2713,7 @@ static void configure_discard_support(struct dm_target *ti, struct raid_set *rs) /* RAID level 4,5,6 require discard_zeroes_data for data integrity! */ raid456 = (rs->md.level == 4 || rs->md.level == 5 || rs->md.level == 6); - for (i = 0; i < rs->md.raid_disks; i++) { + for (i = 0; i < rs->raid_disks; i++) { struct request_queue *q; if (!rs->dev[i].rdev.bdev) @@ -1207,118 +2746,252 @@ static void configure_discard_support(struct dm_target *ti, struct raid_set *rs) } /* - * Construct a RAID4/5/6 mapping: + * Construct a RAID0/1/10/4/5/6 mapping: * Args: - * <raid_type> <#raid_params> <raid_params> \ - * <#raid_devs> { <meta_dev1> <dev1> .. <meta_devN> <devN> } + * <raid_type> <#raid_params> <raid_params>{0,} \ + * <#raid_devs> [<meta_dev1> <dev1>]{1,} * - * <raid_params> varies by <raid_type>. See 'parse_raid_params' for + * <raid_params> varies by <raid_type>. See 'parse_raid_params' for * details on possible <raid_params>. + * + * Userspace is free to initialize the metadata devices, hence the superblocks to + * enforce recreation based on the passed in table parameters. + * */ -static int raid_ctr(struct dm_target *ti, unsigned argc, char **argv) +static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv) { - int ret; + int r; + bool resize; struct raid_type *rt; - unsigned long num_raid_params, num_raid_devs; + unsigned int num_raid_params, num_raid_devs; + sector_t calculated_dev_sectors; struct raid_set *rs = NULL; - - /* Must have at least <raid_type> <#raid_params> */ - if (argc < 2) { - ti->error = "Too few arguments"; + const char *arg; + struct rs_layout rs_layout; + struct dm_arg_set as = { argc, argv }, as_nrd; + struct dm_arg _args[] = { + { 0, as.argc, "Cannot understand number of raid parameters" }, + { 1, 254, "Cannot understand number of raid devices parameters" } + }; + + /* Must have <raid_type> */ + arg = dm_shift_arg(&as); + if (!arg) { + ti->error = "No arguments"; return -EINVAL; } - /* raid type */ - rt = get_raid_type(argv[0]); + rt = get_raid_type(arg); if (!rt) { ti->error = "Unrecognised raid_type"; return -EINVAL; } - argc--; - argv++; - - /* number of RAID parameters */ - if (kstrtoul(argv[0], 10, &num_raid_params) < 0) { - ti->error = "Cannot understand number of RAID parameters"; - return -EINVAL; - } - argc--; - argv++; - /* Skip over RAID params for now and find out # of devices */ - if (num_raid_params >= argc) { - ti->error = "Arguments do not agree with counts given"; + /* Must have <#raid_params> */ + if (dm_read_arg_group(_args, &as, &num_raid_params, &ti->error)) return -EINVAL; - } - if ((kstrtoul(argv[num_raid_params], 10, &num_raid_devs) < 0) || - (num_raid_devs > MAX_RAID_DEVICES)) { - ti->error = "Cannot understand number of raid devices"; + /* number of raid device tupples <meta_dev data_dev> */ + as_nrd = as; + dm_consume_args(&as_nrd, num_raid_params); + _args[1].max = (as_nrd.argc - 1) / 2; + if (dm_read_arg(_args + 1, &as_nrd, &num_raid_devs, &ti->error)) return -EINVAL; - } - argc -= num_raid_params + 1; /* +1: we already have num_raid_devs */ - if (argc != (num_raid_devs * 2)) { - ti->error = "Supplied RAID devices does not match the count given"; + if (!__within_range(num_raid_devs, 1, MAX_RAID_DEVICES)) { + ti->error = "Invalid number of supplied raid devices"; return -EINVAL; } - rs = context_alloc(ti, rt, (unsigned)num_raid_devs); + rs = raid_set_alloc(ti, rt, num_raid_devs); if (IS_ERR(rs)) return PTR_ERR(rs); - ret = parse_raid_params(rs, argv, (unsigned)num_raid_params); - if (ret) + r = parse_raid_params(rs, &as, num_raid_params); + if (r) goto bad; - argv += num_raid_params + 1; - - ret = dev_parms(rs, argv); - if (ret) + r = parse_dev_params(rs, &as); + if (r) goto bad; rs->md.sync_super = super_sync; - ret = analyse_superblocks(ti, rs); - if (ret) + + /* + * Calculate ctr requested array and device sizes to allow + * for superblock analysis needing device sizes defined. + * + * Any existing superblock will overwrite the array and device sizes + */ + r = rs_set_dev_and_array_sectors(rs, false); + if (r) + goto bad; + + calculated_dev_sectors = rs->dev[0].rdev.sectors; + + /* + * Backup any new raid set level, layout, ... + * requested to be able to compare to superblock + * members for conversion decisions. + */ + rs_config_backup(rs, &rs_layout); + + r = analyse_superblocks(ti, rs); + if (r) goto bad; + resize = calculated_dev_sectors != rs->dev[0].rdev.sectors; + INIT_WORK(&rs->md.event_work, do_table_event); ti->private = rs; ti->num_flush_bios = 1; + /* Restore any requested new layout for conversion decision */ + rs_config_restore(rs, &rs_layout); + /* - * Disable/enable discard support on RAID set. + * Now that we have any superblock metadata available, + * check for new, recovering, reshaping, to be taken over, + * to be reshaped or an existing, unchanged raid set to + * run in sequence. */ - configure_discard_support(ti, rs); + if (test_bit(MD_ARRAY_FIRST_USE, &rs->md.flags)) { + /* A new raid6 set has to be recovered to ensure proper parity and Q-Syndrome */ + if (rs_is_raid6(rs) && + test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) { + ti->error = "'nosync' not allowed for new raid6 set"; + r = -EINVAL; + goto bad; + } + rs_setup_recovery(rs, 0); + set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); + rs_set_new(rs); + } else if (rs_is_recovering(rs)) { + /* A recovering raid set may be resized */ + ; /* skip setup rs */ + } else if (rs_is_reshaping(rs)) { + /* Have to reject size change request during reshape */ + if (resize) { + ti->error = "Can't resize a reshaping raid set"; + r = -EPERM; + goto bad; + } + /* skip setup rs */ + } else if (rs_takeover_requested(rs)) { + if (rs_is_reshaping(rs)) { + ti->error = "Can't takeover a reshaping raid set"; + r = -EPERM; + goto bad; + } + + /* + * If a takeover is needed, userspace sets any additional + * devices to rebuild and we can check for a valid request here. + * + * If acceptible, set the level to the new requested + * one, prohibit requesting recovery, allow the raid + * set to run and store superblocks during resume. + */ + r = rs_check_takeover(rs); + if (r) + goto bad; + + r = rs_setup_takeover(rs); + if (r) + goto bad; + + set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); + set_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags); + /* Takeover ain't recovery, so disable recovery */ + rs_setup_recovery(rs, MaxSector); + rs_set_new(rs); + } else if (rs_reshape_requested(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); + rs_set_cur(rs); + } else { + /* May not set recovery when a device rebuild is requested */ + if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags)) { + rs_setup_recovery(rs, MaxSector); + set_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags); + } else + rs_setup_recovery(rs, test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags) ? + 0 : (resize ? calculated_dev_sectors : MaxSector)); + rs_set_cur(rs); + } + + /* If constructor requested it, change data and new_data offsets */ + r = rs_adjust_data_offsets(rs); + if (r) + goto bad; + + /* Start raid set read-only and assumed clean to change in raid_resume() */ + rs->md.ro = 1; + rs->md.in_sync = 1; + set_bit(MD_RECOVERY_FROZEN, &rs->md.recovery); /* Has to be held on running the array */ mddev_lock_nointr(&rs->md); - ret = md_run(&rs->md); + r = md_run(&rs->md); rs->md.in_sync = 0; /* Assume already marked dirty */ - mddev_unlock(&rs->md); - if (ret) { - ti->error = "Fail to run raid array"; + if (r) { + ti->error = "Failed to run raid array"; + mddev_unlock(&rs->md); goto bad; } - if (ti->len != rs->md.array_sectors) { - ti->error = "Array size does not match requested target length"; - ret = -EINVAL; - goto size_mismatch; - } rs->callbacks.congested_fn = raid_is_congested; dm_table_add_target_callbacks(ti->table, &rs->callbacks); mddev_suspend(&rs->md); + + /* Try to adjust the raid4/5/6 stripe cache size to the stripe size */ + if (rs_is_raid456(rs)) { + r = rs_set_raid456_stripe_cache(rs); + if (r) + goto bad_stripe_cache; + } + + /* Now do an early reshape check */ + if (test_bit(RT_FLAG_RESHAPE_RS, &rs->runtime_flags)) { + r = rs_check_reshape(rs); + if (r) + goto bad_check_reshape; + + /* Restore new, ctr requested layout to perform check */ + rs_config_restore(rs, &rs_layout); + + if (rs->md.pers->start_reshape) { + r = rs->md.pers->check_reshape(&rs->md); + if (r) { + ti->error = "Reshape check failed"; + goto bad_check_reshape; + } + } + } + + mddev_unlock(&rs->md); return 0; -size_mismatch: +bad_stripe_cache: +bad_check_reshape: md_stop(&rs->md); bad: - context_free(rs); + raid_set_free(rs); - return ret; + return r; } static void raid_dtr(struct dm_target *ti) @@ -1327,7 +3000,7 @@ static void raid_dtr(struct dm_target *ti) list_del_init(&rs->callbacks.list); md_stop(&rs->md); - context_free(rs); + raid_set_free(rs); } static int raid_map(struct dm_target *ti, struct bio *bio) @@ -1335,11 +3008,23 @@ static int raid_map(struct dm_target *ti, struct bio *bio) struct raid_set *rs = ti->private; struct mddev *mddev = &rs->md; + /* + * If we're reshaping to add disk(s)), ti->len and + * mddev->array_sectors will differ during the process + * (ti->len > mddev->array_sectors), so we have to requeue + * bios with addresses > mddev->array_sectors here or + * there will occur accesses past EOD of the component + * data images thus erroring the raid set. + */ + if (unlikely(bio_end_sector(bio) > mddev->array_sectors)) + return DM_MAPIO_REQUEUE; + mddev->pers->make_request(mddev, bio); return DM_MAPIO_SUBMITTED; } +/* Return string describing the current sync action of @mddev */ static const char *decipher_sync_action(struct mddev *mddev) { if (test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) @@ -1365,195 +3050,260 @@ static const char *decipher_sync_action(struct mddev *mddev) return "idle"; } -static void raid_status(struct dm_target *ti, status_type_t type, - unsigned status_flags, char *result, unsigned maxlen) +/* + * Return status string @rdev + * + * Status characters: + * + * 'D' = Dead/Failed device + * 'a' = Alive but not in-sync + * 'A' = Alive and in-sync + */ +static const char *__raid_dev_status(struct md_rdev *rdev, bool array_in_sync) { - struct raid_set *rs = ti->private; - unsigned raid_param_cnt = 1; /* at least 1 for chunksize */ - unsigned sz = 0; - int i, array_in_sync = 0; - sector_t sync; + if (test_bit(Faulty, &rdev->flags)) + return "D"; + else if (!array_in_sync || !test_bit(In_sync, &rdev->flags)) + return "a"; + else + return "A"; +} - switch (type) { - case STATUSTYPE_INFO: - DMEMIT("%s %d ", rs->raid_type->name, rs->md.raid_disks); +/* 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) +{ + sector_t r, recovery_cp, curr_resync_completed; + struct mddev *mddev = &rs->md; - if (rs->raid_type->level) { - if (test_bit(MD_RECOVERY_RUNNING, &rs->md.recovery)) - sync = rs->md.curr_resync_completed; - else - sync = rs->md.recovery_cp; - - if (sync >= rs->md.resync_max_sectors) { - /* - * Sync complete. - */ - array_in_sync = 1; - sync = rs->md.resync_max_sectors; - } else if (test_bit(MD_RECOVERY_REQUESTED, &rs->md.recovery)) { - /* - * If "check" or "repair" is occurring, the array has - * undergone and initial sync and the health characters - * should not be 'a' anymore. - */ - array_in_sync = 1; + curr_resync_completed = mddev->curr_resync_completed ?: mddev->recovery_cp; + recovery_cp = mddev->recovery_cp; + *array_in_sync = false; + + if (rs_is_raid0(rs)) { + r = resync_max_sectors; + *array_in_sync = true; + + } 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 { - /* - * The array may be doing an initial sync, or it may - * be rebuilding individual components. If all the - * devices are In_sync, then it is the array that is - * being initialized. - */ - for (i = 0; i < rs->md.raid_disks; i++) - if (!test_bit(In_sync, &rs->dev[i].rdev.flags)) - array_in_sync = 1; + /* 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; + else + r = recovery_cp; + + if (r == MaxSector) { + /* + * Sync complete. + */ + *array_in_sync = true; + r = resync_max_sectors; + } else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->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; } else { - /* RAID0 */ - array_in_sync = 1; - sync = rs->md.resync_max_sectors; - } + struct md_rdev *rdev; - /* - * Status characters: - * 'D' = Dead/Failed device - * 'a' = Alive but not in-sync - * 'A' = Alive and in-sync - */ - for (i = 0; i < rs->md.raid_disks; i++) { - if (test_bit(Faulty, &rs->dev[i].rdev.flags)) - DMEMIT("D"); - else if (!array_in_sync || - !test_bit(In_sync, &rs->dev[i].rdev.flags)) - DMEMIT("a"); - else - DMEMIT("A"); + /* + * 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. + */ + rdev_for_each(rdev, mddev) + if (!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 } + } + + return r; +} + +/* Helper to return @dev name or "-" if !@dev */ +static const char *__get_dev_name(struct dm_dev *dev) +{ + return dev ? dev->name : "-"; +} + +static void raid_status(struct dm_target *ti, status_type_t type, + unsigned int status_flags, char *result, unsigned int maxlen) +{ + struct raid_set *rs = ti->private; + 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 int raid_param_cnt = 1; /* at least 1 for chunksize */ + unsigned int sz = 0; + unsigned int rebuild_disks; + unsigned int write_mostly_params = 0; + sector_t progress, resync_max_sectors, resync_mismatches; + const char *sync_action; + struct raid_type *rt; + struct md_rdev *rdev; + + switch (type) { + case STATUSTYPE_INFO: + /* *Should* always succeed */ + rt = get_raid_type_by_ll(mddev->new_level, mddev->new_layout); + if (!rt) + return; + + DMEMIT("%s %d ", rt->name, mddev->raid_disks); + + /* Access most recent mddev properties for status output */ + smp_rmb(); + /* 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); + 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); + + /* HM FIXME: do we want another state char for raid0? It shows 'D' or 'A' now */ + rdev_for_each(rdev, mddev) + DMEMIT(__raid_dev_status(rdev, array_in_sync)); /* - * In-sync ratio: + * In-sync/Reshape ratio: * The in-sync ratio shows the progress of: - * - Initializing the array - * - Rebuilding a subset of devices of the array + * - Initializing the raid set + * - Rebuilding a subset of devices of the raid set * The user can distinguish between the two by referring * to the status characters. + * + * The reshape ratio shows the progress of + * changing the raid layout or the number of + * disks of a raid set */ - DMEMIT(" %llu/%llu", - (unsigned long long) sync, - (unsigned long long) rs->md.resync_max_sectors); + DMEMIT(" %llu/%llu", (unsigned long long) progress, + (unsigned long long) resync_max_sectors); /* + * v1.5.0+: + * * Sync action: - * See Documentation/device-mapper/dm-raid.c for + * See Documentation/device-mapper/dm-raid.txt for * information on each of these states. */ - DMEMIT(" %s", decipher_sync_action(&rs->md)); + DMEMIT(" %s", sync_action); /* + * v1.5.0+: + * * resync_mismatches/mismatch_cnt * This field shows the number of discrepancies found when - * performing a "check" of the array. + * performing a "check" of the raid set. */ - DMEMIT(" %llu", - (strcmp(rs->md.last_sync_action, "check")) ? 0 : - (unsigned long long) - atomic64_read(&rs->md.resync_mismatches)); - break; - case STATUSTYPE_TABLE: - /* The string you would use to construct this array */ - for (i = 0; i < rs->md.raid_disks; i++) { - if ((rs->ctr_flags & CTR_FLAG_REBUILD) && - rs->dev[i].data_dev && - !test_bit(In_sync, &rs->dev[i].rdev.flags)) - raid_param_cnt += 2; /* for rebuilds */ - if (rs->dev[i].data_dev && - test_bit(WriteMostly, &rs->dev[i].rdev.flags)) - raid_param_cnt += 2; - } - - raid_param_cnt += (hweight32(rs->ctr_flags & ~CTR_FLAG_REBUILD) * 2); - if (rs->ctr_flags & (CTR_FLAG_SYNC | CTR_FLAG_NOSYNC)) - raid_param_cnt--; - - DMEMIT("%s %u %u", rs->raid_type->name, - raid_param_cnt, rs->md.chunk_sectors); - - if ((rs->ctr_flags & CTR_FLAG_SYNC) && - (rs->md.recovery_cp == MaxSector)) - DMEMIT(" sync"); - if (rs->ctr_flags & CTR_FLAG_NOSYNC) - DMEMIT(" nosync"); - - for (i = 0; i < rs->md.raid_disks; i++) - if ((rs->ctr_flags & CTR_FLAG_REBUILD) && - rs->dev[i].data_dev && - !test_bit(In_sync, &rs->dev[i].rdev.flags)) - DMEMIT(" rebuild %u", i); - - if (rs->ctr_flags & CTR_FLAG_DAEMON_SLEEP) - DMEMIT(" daemon_sleep %lu", - rs->md.bitmap_info.daemon_sleep); - - if (rs->ctr_flags & CTR_FLAG_MIN_RECOVERY_RATE) - DMEMIT(" min_recovery_rate %d", rs->md.sync_speed_min); - - if (rs->ctr_flags & CTR_FLAG_MAX_RECOVERY_RATE) - DMEMIT(" max_recovery_rate %d", rs->md.sync_speed_max); - - for (i = 0; i < rs->md.raid_disks; i++) - if (rs->dev[i].data_dev && - test_bit(WriteMostly, &rs->dev[i].rdev.flags)) - DMEMIT(" write_mostly %u", i); - - if (rs->ctr_flags & CTR_FLAG_MAX_WRITE_BEHIND) - DMEMIT(" max_write_behind %lu", - rs->md.bitmap_info.max_write_behind); - - if (rs->ctr_flags & CTR_FLAG_STRIPE_CACHE) { - struct r5conf *conf = rs->md.private; - - /* convert from kiB to sectors */ - DMEMIT(" stripe_cache %d", - conf ? conf->max_nr_stripes * 2 : 0); - } - - if (rs->ctr_flags & CTR_FLAG_REGION_SIZE) - DMEMIT(" region_size %lu", - rs->md.bitmap_info.chunksize >> 9); - - if (rs->ctr_flags & CTR_FLAG_RAID10_COPIES) - DMEMIT(" raid10_copies %u", - raid10_md_layout_to_copies(rs->md.layout)); - - if (rs->ctr_flags & CTR_FLAG_RAID10_FORMAT) - DMEMIT(" raid10_format %s", - raid10_md_layout_to_format(rs->md.layout)); + DMEMIT(" %llu", (unsigned long long) resync_mismatches); - DMEMIT(" %d", rs->md.raid_disks); - for (i = 0; i < rs->md.raid_disks; i++) { - if (rs->dev[i].meta_dev) - DMEMIT(" %s", rs->dev[i].meta_dev->name); - else - DMEMIT(" -"); + /* + * v1.9.0+: + * + * data_offset (needed for out of space reshaping) + * This field shows the data offset into the data + * image LV where the first stripes data starts. + * + * We keep data_offset equal on all raid disks of the set, + * so retrieving it from the first raid disk is sufficient. + */ + DMEMIT(" %llu", (unsigned long long) rs->dev[0].rdev.data_offset); + break; - if (rs->dev[i].data_dev) - DMEMIT(" %s", rs->dev[i].data_dev->name); - else - DMEMIT(" -"); - } + case STATUSTYPE_TABLE: + /* Report the table line string you would use to construct this raid set */ + + /* Calculate raid parameter count */ + for (i = 0; i < rs->raid_disks; i++) + if (test_bit(WriteMostly, &rs->dev[i].rdev.flags)) + write_mostly_params += 2; + rebuild_disks = memweight(rs->rebuild_disks, DISKS_ARRAY_ELEMS * sizeof(*rs->rebuild_disks)); + raid_param_cnt += rebuild_disks * 2 + + write_mostly_params + + hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_NO_ARGS) + + hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_ONE_ARG) * 2; + /* Emit table line */ + DMEMIT("%s %u %u", rs->raid_type->name, raid_param_cnt, mddev->new_chunk_sectors); + if (test_bit(__CTR_FLAG_RAID10_FORMAT, &rs->ctr_flags)) + DMEMIT(" %s %s", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_FORMAT), + raid10_md_layout_to_format(mddev->layout)); + if (test_bit(__CTR_FLAG_RAID10_COPIES, &rs->ctr_flags)) + DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_RAID10_COPIES), + raid10_md_layout_to_copies(mddev->layout)); + if (test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) + DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC)); + if (test_bit(__CTR_FLAG_SYNC, &rs->ctr_flags)) + DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_SYNC)); + if (test_bit(__CTR_FLAG_REGION_SIZE, &rs->ctr_flags)) + DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_REGION_SIZE), + (unsigned long long) to_sector(mddev->bitmap_info.chunksize)); + if (test_bit(__CTR_FLAG_DATA_OFFSET, &rs->ctr_flags)) + DMEMIT(" %s %llu", dm_raid_arg_name_by_flag(CTR_FLAG_DATA_OFFSET), + (unsigned long long) rs->data_offset); + if (test_bit(__CTR_FLAG_DAEMON_SLEEP, &rs->ctr_flags)) + DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_DAEMON_SLEEP), + mddev->bitmap_info.daemon_sleep); + if (test_bit(__CTR_FLAG_DELTA_DISKS, &rs->ctr_flags)) + DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_DELTA_DISKS), + max(rs->delta_disks, mddev->delta_disks)); + if (test_bit(__CTR_FLAG_STRIPE_CACHE, &rs->ctr_flags)) + DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_STRIPE_CACHE), + max_nr_stripes); + if (rebuild_disks) + for (i = 0; i < rs->raid_disks; i++) + if (test_bit(rs->dev[i].rdev.raid_disk, (void *) rs->rebuild_disks)) + DMEMIT(" %s %u", dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD), + rs->dev[i].rdev.raid_disk); + if (write_mostly_params) + for (i = 0; i < rs->raid_disks; i++) + if (test_bit(WriteMostly, &rs->dev[i].rdev.flags)) + DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_WRITE_MOSTLY), + rs->dev[i].rdev.raid_disk); + if (test_bit(__CTR_FLAG_MAX_WRITE_BEHIND, &rs->ctr_flags)) + DMEMIT(" %s %lu", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_WRITE_BEHIND), + mddev->bitmap_info.max_write_behind); + if (test_bit(__CTR_FLAG_MAX_RECOVERY_RATE, &rs->ctr_flags)) + DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MAX_RECOVERY_RATE), + mddev->sync_speed_max); + if (test_bit(__CTR_FLAG_MIN_RECOVERY_RATE, &rs->ctr_flags)) + DMEMIT(" %s %d", dm_raid_arg_name_by_flag(CTR_FLAG_MIN_RECOVERY_RATE), + mddev->sync_speed_min); + DMEMIT(" %d", rs->raid_disks); + for (i = 0; i < rs->raid_disks; i++) + DMEMIT(" %s %s", __get_dev_name(rs->dev[i].meta_dev), + __get_dev_name(rs->dev[i].data_dev)); } } -static int raid_message(struct dm_target *ti, unsigned argc, char **argv) +static int raid_message(struct dm_target *ti, unsigned int argc, char **argv) { struct raid_set *rs = ti->private; struct mddev *mddev = &rs->md; - if (!strcasecmp(argv[0], "reshape")) { - DMERR("Reshape not supported."); - return -EINVAL; - } - if (!mddev->pers || !mddev->pers->sync_request) return -EINVAL; @@ -1571,11 +3321,10 @@ static int raid_message(struct dm_target *ti, unsigned argc, char **argv) test_bit(MD_RECOVERY_NEEDED, &mddev->recovery)) return -EBUSY; else if (!strcasecmp(argv[0], "resync")) - set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); - else if (!strcasecmp(argv[0], "recover")) { + ; /* MD_RECOVERY_NEEDED set below */ + else if (!strcasecmp(argv[0], "recover")) set_bit(MD_RECOVERY_RECOVER, &mddev->recovery); - set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); - } else { + else { if (!strcasecmp(argv[0], "check")) set_bit(MD_RECOVERY_CHECK, &mddev->recovery); else if (!!strcasecmp(argv[0], "repair")) @@ -1588,11 +3337,11 @@ static int raid_message(struct dm_target *ti, unsigned argc, char **argv) * canceling read-auto mode */ mddev->ro = 0; - if (!mddev->suspended) + if (!mddev->suspended && mddev->sync_thread) md_wakeup_thread(mddev->sync_thread); } set_bit(MD_RECOVERY_NEEDED, &mddev->recovery); - if (!mddev->suspended) + if (!mddev->suspended && mddev->thread) md_wakeup_thread(mddev->thread); return 0; @@ -1602,28 +3351,27 @@ static int raid_iterate_devices(struct dm_target *ti, iterate_devices_callout_fn fn, void *data) { struct raid_set *rs = ti->private; - unsigned i; - int ret = 0; + unsigned int i; + int r = 0; - for (i = 0; !ret && i < rs->md.raid_disks; i++) + for (i = 0; !r && i < rs->md.raid_disks; i++) if (rs->dev[i].data_dev) - ret = fn(ti, + r = fn(ti, rs->dev[i].data_dev, 0, /* No offset on data devs */ rs->md.dev_sectors, data); - return ret; + return r; } static void raid_io_hints(struct dm_target *ti, struct queue_limits *limits) { struct raid_set *rs = ti->private; - unsigned chunk_size = rs->md.chunk_sectors << 9; - struct r5conf *conf = rs->md.private; + unsigned int chunk_size = to_bytes(rs->md.chunk_sectors); blk_limits_io_min(limits, chunk_size); - blk_limits_io_opt(limits, chunk_size * (conf->raid_disks - conf->max_degraded)); + blk_limits_io_opt(limits, chunk_size * mddev_data_stripes(rs)); } static void raid_presuspend(struct dm_target *ti) @@ -1637,7 +3385,11 @@ static void raid_postsuspend(struct dm_target *ti) { struct raid_set *rs = ti->private; - mddev_suspend(&rs->md); + if (test_and_clear_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) { + if (!rs->md.suspended) + mddev_suspend(&rs->md); + rs->md.ro = 1; + } } static void attempt_restore_of_faulty_devices(struct raid_set *rs) @@ -1651,8 +3403,8 @@ static void attempt_restore_of_faulty_devices(struct raid_set *rs) for (i = 0; i < rs->md.raid_disks; i++) { r = &rs->dev[i].rdev; if (test_bit(Faulty, &r->flags) && r->sb_page && - sync_page_io(r, 0, r->sb_size, r->sb_page, REQ_OP_READ, 0, - 1)) { + sync_page_io(r, 0, r->sb_size, r->sb_page, + REQ_OP_READ, 0, true)) { DMINFO("Faulty %s device #%d has readable super block." " Attempting to revive it.", rs->raid_type->name, i); @@ -1661,7 +3413,7 @@ static void attempt_restore_of_faulty_devices(struct raid_set *rs) * Faulty bit may be set, but sometimes the array can * be suspended before the personalities can respond * by removing the device from the array (i.e. calling - * 'hot_remove_disk'). If they haven't yet removed + * 'hot_remove_disk'). If they haven't yet removed * the failed device, its 'raid_disk' number will be * '>= 0' - meaning we must call this function * ourselves. @@ -1697,34 +3449,192 @@ static void attempt_restore_of_faulty_devices(struct raid_set *rs) } } -static void raid_resume(struct dm_target *ti) +static int __load_dirty_region_bitmap(struct raid_set *rs) { - struct raid_set *rs = ti->private; + int r = 0; + + /* Try loading the bitmap unless "raid0", which does not have one */ + if (!rs_is_raid0(rs) && + !test_and_set_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags)) { + r = bitmap_load(&rs->md); + if (r) + DMERR("Failed to load bitmap"); + } - if (rs->raid_type->level) { - set_bit(MD_CHANGE_DEVS, &rs->md.flags); + return r; +} - if (!rs->bitmap_loaded) { - bitmap_load(&rs->md); - rs->bitmap_loaded = 1; - } else { - /* - * A secondary resume while the device is active. - * Take this opportunity to check whether any failed - * devices are reachable again. - */ - attempt_restore_of_faulty_devices(rs); +/* Enforce updating all superblocks */ +static void rs_update_sbs(struct raid_set *rs) +{ + struct mddev *mddev = &rs->md; + int ro = mddev->ro; + + set_bit(MD_CHANGE_DEVS, &mddev->flags); + mddev->ro = 0; + md_update_sb(mddev, 1); + mddev->ro = ro; +} + +/* + * Reshape changes raid algorithm of @rs to new one within personality + * (e.g. raid6_zr -> raid6_nc), changes stripe size, adds/removes + * disks from a raid set thus growing/shrinking it or resizes the set + * + * Call mddev_lock_nointr() before! + */ +static int rs_start_reshape(struct raid_set *rs) +{ + int r; + struct mddev *mddev = &rs->md; + struct md_personality *pers = mddev->pers; + + r = rs_setup_reshape(rs); + if (r) + return r; + + /* Need to be resumed to be able to start reshape, recovery is frozen until raid_resume() though */ + if (mddev->suspended) + mddev_resume(mddev); + + /* + * Check any reshape constraints enforced by the personalility + * + * May as well already kick the reshape off so that * pers->start_reshape() becomes optional. + */ + r = pers->check_reshape(mddev); + if (r) { + rs->ti->error = "pers->check_reshape() failed"; + return r; + } + + /* + * Personality may not provide start reshape method in which + * case check_reshape above has already covered everything + */ + if (pers->start_reshape) { + r = pers->start_reshape(mddev); + if (r) { + rs->ti->error = "pers->start_reshape() failed"; + return r; } + } + + /* Suspend because a resume will happen in raid_resume() */ + if (!mddev->suspended) + mddev_suspend(mddev); + + /* + * Now reshape got set up, update superblocks to + * reflect the fact so that a table reload will + * access proper superblock content in the ctr. + */ + rs_update_sbs(rs); + + return 0; +} + +static int raid_preresume(struct dm_target *ti) +{ + int r; + 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 */ + if (test_and_set_bit(RT_FLAG_RS_PRERESUMED, &rs->runtime_flags)) + return 0; + + /* + * The superblocks need to be updated on disk if the + * array is new or new devices got added (thus zeroed + * out by userspace) or __load_dirty_region_bitmap + * will overwrite them in core with old data or fail. + */ + if (test_bit(RT_FLAG_UPDATE_SBS, &rs->runtime_flags)) + rs_update_sbs(rs); - clear_bit(MD_RECOVERY_FROZEN, &rs->md.recovery); + /* + * Disable/enable discard support on raid set after any + * conversion, because devices can have been added + */ + configure_discard_support(rs); + + /* Load the bitmap from disk unless raid0 */ + r = __load_dirty_region_bitmap(rs); + if (r) + return r; + + /* Resize bitmap to adjust to changed region size (aka MD bitmap chunksize) */ + if (test_bit(RT_FLAG_RS_BITMAP_LOADED, &rs->runtime_flags) && + mddev->bitmap_info.chunksize != to_bytes(rs->requested_bitmap_chunk_sectors)) { + r = bitmap_resize(mddev->bitmap, mddev->dev_sectors, + to_bytes(rs->requested_bitmap_chunk_sectors), 0); + if (r) + DMERR("Failed to resize bitmap"); + } + + /* Check for any resize/reshape on @rs and adjust/initiate */ + /* Be prepared for mddev_resume() in raid_resume() */ + set_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + if (mddev->recovery_cp && mddev->recovery_cp < MaxSector) { + set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery); + set_bit(MD_RECOVERY_SYNC, &mddev->recovery); + mddev->resync_min = mddev->recovery_cp; } - mddev_resume(&rs->md); + 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)) { + /* Initiate a reshape. */ + mddev_lock_nointr(mddev); + r = rs_start_reshape(rs); + mddev_unlock(mddev); + if (r) + DMWARN("Failed to check/start reshape, continuing without change"); + r = 0; + } + + return r; +} + +static void raid_resume(struct dm_target *ti) +{ + struct raid_set *rs = ti->private; + struct mddev *mddev = &rs->md; + + if (test_and_set_bit(RT_FLAG_RS_RESUMED, &rs->runtime_flags)) { + /* + * A secondary resume while the device is active. + * Take this opportunity to check whether any failed + * devices are reachable again. + */ + attempt_restore_of_faulty_devices(rs); + } else { + mddev->ro = 0; + mddev->in_sync = 0; + + /* + * When passing in flags to the ctr, we expect userspace + * to reset them because they made it to the superblocks + * and reload the mapping anyway. + * + * -> only unfreeze recovery in case of a table reload or + * we'll have a bogus recovery/reshape position + * retrieved from the superblock by the ctr because + * the ongoing recovery/reshape will change it after read. + */ + if (!test_bit(RT_FLAG_KEEP_RS_FROZEN, &rs->runtime_flags)) + clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery); + + if (mddev->suspended) + mddev_resume(mddev); + } } static struct target_type raid_target = { .name = "raid", - .version = {1, 8, 0}, + .version = {1, 9, 0}, .module = THIS_MODULE, .ctr = raid_ctr, .dtr = raid_dtr, @@ -1735,6 +3645,7 @@ static struct target_type raid_target = { .io_hints = raid_io_hints, .presuspend = raid_presuspend, .postsuspend = raid_postsuspend, + .preresume = raid_preresume, .resume = raid_resume, }; @@ -1759,11 +3670,13 @@ module_param(devices_handle_discard_safely, bool, 0644); MODULE_PARM_DESC(devices_handle_discard_safely, "Set to Y if all devices in each array reliably return zeroes on reads from discarded regions"); -MODULE_DESCRIPTION(DM_NAME " raid4/5/6 target"); +MODULE_DESCRIPTION(DM_NAME " raid0/1/10/4/5/6 target"); +MODULE_ALIAS("dm-raid0"); MODULE_ALIAS("dm-raid1"); MODULE_ALIAS("dm-raid10"); MODULE_ALIAS("dm-raid4"); MODULE_ALIAS("dm-raid5"); MODULE_ALIAS("dm-raid6"); MODULE_AUTHOR("Neil Brown <dm-devel@redhat.com>"); +MODULE_AUTHOR("Heinz Mauelshagen <dm-devel@redhat.com>"); MODULE_LICENSE("GPL"); diff --git a/drivers/md/dm-rq.c b/drivers/md/dm-rq.c new file mode 100644 index 000000000000..7a9661868496 --- /dev/null +++ b/drivers/md/dm-rq.c @@ -0,0 +1,970 @@ +/* + * Copyright (C) 2016 Red Hat, Inc. All rights reserved. + * + * This file is released under the GPL. + */ + +#include "dm-core.h" +#include "dm-rq.h" + +#include <linux/elevator.h> /* for rq_end_sector() */ +#include <linux/blk-mq.h> + +#define DM_MSG_PREFIX "core-rq" + +#define DM_MQ_NR_HW_QUEUES 1 +#define DM_MQ_QUEUE_DEPTH 2048 +static unsigned dm_mq_nr_hw_queues = DM_MQ_NR_HW_QUEUES; +static unsigned dm_mq_queue_depth = DM_MQ_QUEUE_DEPTH; + +/* + * Request-based DM's mempools' reserved IOs set by the user. + */ +#define RESERVED_REQUEST_BASED_IOS 256 +static unsigned reserved_rq_based_ios = RESERVED_REQUEST_BASED_IOS; + +#ifdef CONFIG_DM_MQ_DEFAULT +static bool use_blk_mq = true; +#else +static bool use_blk_mq = false; +#endif + +bool dm_use_blk_mq_default(void) +{ + return use_blk_mq; +} + +bool dm_use_blk_mq(struct mapped_device *md) +{ + return md->use_blk_mq; +} +EXPORT_SYMBOL_GPL(dm_use_blk_mq); + +unsigned dm_get_reserved_rq_based_ios(void) +{ + return __dm_get_module_param(&reserved_rq_based_ios, + RESERVED_REQUEST_BASED_IOS, DM_RESERVED_MAX_IOS); +} +EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios); + +static unsigned dm_get_blk_mq_nr_hw_queues(void) +{ + return __dm_get_module_param(&dm_mq_nr_hw_queues, 1, 32); +} + +static unsigned dm_get_blk_mq_queue_depth(void) +{ + return __dm_get_module_param(&dm_mq_queue_depth, + DM_MQ_QUEUE_DEPTH, BLK_MQ_MAX_DEPTH); +} + +int dm_request_based(struct mapped_device *md) +{ + return blk_queue_stackable(md->queue); +} + +static void dm_old_start_queue(struct request_queue *q) +{ + unsigned long flags; + + spin_lock_irqsave(q->queue_lock, flags); + if (blk_queue_stopped(q)) + blk_start_queue(q); + spin_unlock_irqrestore(q->queue_lock, flags); +} + +void dm_start_queue(struct request_queue *q) +{ + if (!q->mq_ops) + dm_old_start_queue(q); + else { + blk_mq_start_stopped_hw_queues(q, true); + blk_mq_kick_requeue_list(q); + } +} + +static void dm_old_stop_queue(struct request_queue *q) +{ + unsigned long flags; + + spin_lock_irqsave(q->queue_lock, flags); + if (blk_queue_stopped(q)) { + spin_unlock_irqrestore(q->queue_lock, flags); + return; + } + + blk_stop_queue(q); + spin_unlock_irqrestore(q->queue_lock, flags); +} + +void dm_stop_queue(struct request_queue *q) +{ + if (!q->mq_ops) + dm_old_stop_queue(q); + else + blk_mq_stop_hw_queues(q); +} + +static struct dm_rq_target_io *alloc_old_rq_tio(struct mapped_device *md, + gfp_t gfp_mask) +{ + return mempool_alloc(md->io_pool, gfp_mask); +} + +static void free_old_rq_tio(struct dm_rq_target_io *tio) +{ + mempool_free(tio, tio->md->io_pool); +} + +static struct request *alloc_old_clone_request(struct mapped_device *md, + gfp_t gfp_mask) +{ + return mempool_alloc(md->rq_pool, gfp_mask); +} + +static void free_old_clone_request(struct mapped_device *md, struct request *rq) +{ + mempool_free(rq, md->rq_pool); +} + +/* + * Partial completion handling for request-based dm + */ +static void end_clone_bio(struct bio *clone) +{ + struct dm_rq_clone_bio_info *info = + container_of(clone, struct dm_rq_clone_bio_info, clone); + struct dm_rq_target_io *tio = info->tio; + struct bio *bio = info->orig; + unsigned int nr_bytes = info->orig->bi_iter.bi_size; + int error = clone->bi_error; + + bio_put(clone); + + if (tio->error) + /* + * An error has already been detected on the request. + * Once error occurred, just let clone->end_io() handle + * the remainder. + */ + return; + else if (error) { + /* + * Don't notice the error to the upper layer yet. + * The error handling decision is made by the target driver, + * when the request is completed. + */ + tio->error = error; + return; + } + + /* + * I/O for the bio successfully completed. + * Notice the data completion to the upper layer. + */ + + /* + * bios are processed from the head of the list. + * So the completing bio should always be rq->bio. + * If it's not, something wrong is happening. + */ + if (tio->orig->bio != bio) + DMERR("bio completion is going in the middle of the request"); + + /* + * Update the original request. + * Do not use blk_end_request() here, because it may complete + * the original request before the clone, and break the ordering. + */ + blk_update_request(tio->orig, 0, nr_bytes); +} + +static struct dm_rq_target_io *tio_from_request(struct request *rq) +{ + return (rq->q->mq_ops ? blk_mq_rq_to_pdu(rq) : rq->special); +} + +static void rq_end_stats(struct mapped_device *md, struct request *orig) +{ + if (unlikely(dm_stats_used(&md->stats))) { + struct dm_rq_target_io *tio = tio_from_request(orig); + tio->duration_jiffies = jiffies - tio->duration_jiffies; + dm_stats_account_io(&md->stats, rq_data_dir(orig), + blk_rq_pos(orig), tio->n_sectors, true, + tio->duration_jiffies, &tio->stats_aux); + } +} + +/* + * Don't touch any member of the md after calling this function because + * the md may be freed in dm_put() at the end of this function. + * Or do dm_get() before calling this function and dm_put() later. + */ +static void rq_completed(struct mapped_device *md, int rw, bool run_queue) +{ + atomic_dec(&md->pending[rw]); + + /* nudge anyone waiting on suspend queue */ + if (!md_in_flight(md)) + wake_up(&md->wait); + + /* + * Run this off this callpath, as drivers could invoke end_io while + * inside their request_fn (and holding the queue lock). Calling + * back into ->request_fn() could deadlock attempting to grab the + * queue lock again. + */ + if (!md->queue->mq_ops && run_queue) + blk_run_queue_async(md->queue); + + /* + * dm_put() must be at the end of this function. See the comment above + */ + dm_put(md); +} + +static void free_rq_clone(struct request *clone) +{ + struct dm_rq_target_io *tio = clone->end_io_data; + struct mapped_device *md = tio->md; + + blk_rq_unprep_clone(clone); + + /* + * It is possible for a clone_old_rq() allocated clone to + * get passed in -- it may not yet have a request_queue. + * This is known to occur if the error target replaces + * a multipath target that has a request_fn queue stacked + * on blk-mq queue(s). + */ + if (clone->q && clone->q->mq_ops) + /* stacked on blk-mq queue(s) */ + tio->ti->type->release_clone_rq(clone); + else if (!md->queue->mq_ops) + /* request_fn queue stacked on request_fn queue(s) */ + free_old_clone_request(md, clone); + + if (!md->queue->mq_ops) + free_old_rq_tio(tio); +} + +/* + * Complete the clone and the original request. + * Must be called without clone's queue lock held, + * see end_clone_request() for more details. + */ +static void dm_end_request(struct request *clone, int error) +{ + int rw = rq_data_dir(clone); + struct dm_rq_target_io *tio = clone->end_io_data; + struct mapped_device *md = tio->md; + struct request *rq = tio->orig; + + if (rq->cmd_type == REQ_TYPE_BLOCK_PC) { + rq->errors = clone->errors; + rq->resid_len = clone->resid_len; + + if (rq->sense) + /* + * We are using the sense buffer of the original + * request. + * So setting the length of the sense data is enough. + */ + rq->sense_len = clone->sense_len; + } + + free_rq_clone(clone); + rq_end_stats(md, rq); + if (!rq->q->mq_ops) + blk_end_request_all(rq, error); + else + blk_mq_end_request(rq, error); + rq_completed(md, rw, true); +} + +static void dm_unprep_request(struct request *rq) +{ + struct dm_rq_target_io *tio = tio_from_request(rq); + struct request *clone = tio->clone; + + if (!rq->q->mq_ops) { + rq->special = NULL; + rq->cmd_flags &= ~REQ_DONTPREP; + } + + if (clone) + free_rq_clone(clone); + else if (!tio->md->queue->mq_ops) + free_old_rq_tio(tio); +} + +/* + * Requeue the original request of a clone. + */ +static void dm_old_requeue_request(struct request *rq) +{ + struct request_queue *q = rq->q; + unsigned long flags; + + spin_lock_irqsave(q->queue_lock, flags); + blk_requeue_request(q, rq); + blk_run_queue_async(q); + spin_unlock_irqrestore(q->queue_lock, flags); +} + +static void dm_mq_requeue_request(struct request *rq) +{ + struct request_queue *q = rq->q; + unsigned long flags; + + blk_mq_requeue_request(rq); + spin_lock_irqsave(q->queue_lock, flags); + if (!blk_queue_stopped(q)) + blk_mq_kick_requeue_list(q); + spin_unlock_irqrestore(q->queue_lock, flags); +} + +static void dm_requeue_original_request(struct mapped_device *md, + struct request *rq) +{ + int rw = rq_data_dir(rq); + + rq_end_stats(md, rq); + dm_unprep_request(rq); + + if (!rq->q->mq_ops) + dm_old_requeue_request(rq); + else + dm_mq_requeue_request(rq); + + rq_completed(md, rw, false); +} + +static void dm_done(struct request *clone, int error, bool mapped) +{ + int r = error; + struct dm_rq_target_io *tio = clone->end_io_data; + dm_request_endio_fn rq_end_io = NULL; + + if (tio->ti) { + rq_end_io = tio->ti->type->rq_end_io; + + if (mapped && rq_end_io) + r = rq_end_io(tio->ti, clone, error, &tio->info); + } + + if (unlikely(r == -EREMOTEIO && (req_op(clone) == REQ_OP_WRITE_SAME) && + !clone->q->limits.max_write_same_sectors)) + disable_write_same(tio->md); + + if (r <= 0) + /* The target wants to complete the I/O */ + dm_end_request(clone, r); + else if (r == DM_ENDIO_INCOMPLETE) + /* The target will handle the I/O */ + return; + else if (r == DM_ENDIO_REQUEUE) + /* The target wants to requeue the I/O */ + dm_requeue_original_request(tio->md, tio->orig); + else { + DMWARN("unimplemented target endio return value: %d", r); + BUG(); + } +} + +/* + * Request completion handler for request-based dm + */ +static void dm_softirq_done(struct request *rq) +{ + bool mapped = true; + struct dm_rq_target_io *tio = tio_from_request(rq); + struct request *clone = tio->clone; + int rw; + + if (!clone) { + rq_end_stats(tio->md, rq); + rw = rq_data_dir(rq); + if (!rq->q->mq_ops) { + blk_end_request_all(rq, tio->error); + rq_completed(tio->md, rw, false); + free_old_rq_tio(tio); + } else { + blk_mq_end_request(rq, tio->error); + rq_completed(tio->md, rw, false); + } + return; + } + + if (rq->cmd_flags & REQ_FAILED) + mapped = false; + + dm_done(clone, tio->error, mapped); +} + +/* + * Complete the clone and the original request with the error status + * through softirq context. + */ +static void dm_complete_request(struct request *rq, int error) +{ + struct dm_rq_target_io *tio = tio_from_request(rq); + + tio->error = error; + if (!rq->q->mq_ops) + blk_complete_request(rq); + else + blk_mq_complete_request(rq, error); +} + +/* + * Complete the not-mapped clone and the original request with the error status + * through softirq context. + * Target's rq_end_io() function isn't called. + * This may be used when the target's map_rq() or clone_and_map_rq() functions fail. + */ +static void dm_kill_unmapped_request(struct request *rq, int error) +{ + rq->cmd_flags |= REQ_FAILED; + dm_complete_request(rq, error); +} + +/* + * Called with the clone's queue lock held (in the case of .request_fn) + */ +static void end_clone_request(struct request *clone, int error) +{ + struct dm_rq_target_io *tio = clone->end_io_data; + + if (!clone->q->mq_ops) { + /* + * For just cleaning up the information of the queue in which + * the clone was dispatched. + * The clone is *NOT* freed actually here because it is alloced + * from dm own mempool (REQ_ALLOCED isn't set). + */ + __blk_put_request(clone->q, clone); + } + + /* + * Actual request completion is done in a softirq context which doesn't + * hold the clone's queue lock. Otherwise, deadlock could occur because: + * - another request may be submitted by the upper level driver + * of the stacking during the completion + * - the submission which requires queue lock may be done + * against this clone's queue + */ + dm_complete_request(tio->orig, error); +} + +static void dm_dispatch_clone_request(struct request *clone, struct request *rq) +{ + int r; + + if (blk_queue_io_stat(clone->q)) + clone->cmd_flags |= REQ_IO_STAT; + + clone->start_time = jiffies; + r = blk_insert_cloned_request(clone->q, clone); + if (r) + /* must complete clone in terms of original request */ + dm_complete_request(rq, r); +} + +static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig, + void *data) +{ + struct dm_rq_target_io *tio = data; + struct dm_rq_clone_bio_info *info = + container_of(bio, struct dm_rq_clone_bio_info, clone); + + info->orig = bio_orig; + info->tio = tio; + bio->bi_end_io = end_clone_bio; + + return 0; +} + +static int setup_clone(struct request *clone, struct request *rq, + struct dm_rq_target_io *tio, gfp_t gfp_mask) +{ + int r; + + r = blk_rq_prep_clone(clone, rq, tio->md->bs, gfp_mask, + dm_rq_bio_constructor, tio); + if (r) + return r; + + clone->cmd = rq->cmd; + clone->cmd_len = rq->cmd_len; + clone->sense = rq->sense; + clone->end_io = end_clone_request; + clone->end_io_data = tio; + + tio->clone = clone; + + return 0; +} + +static struct request *clone_old_rq(struct request *rq, struct mapped_device *md, + struct dm_rq_target_io *tio, gfp_t gfp_mask) +{ + /* + * Create clone for use with .request_fn request_queue + */ + struct request *clone; + + clone = alloc_old_clone_request(md, gfp_mask); + if (!clone) + return NULL; + + blk_rq_init(NULL, clone); + if (setup_clone(clone, rq, tio, gfp_mask)) { + /* -ENOMEM */ + free_old_clone_request(md, clone); + return NULL; + } + + return clone; +} + +static void map_tio_request(struct kthread_work *work); + +static void init_tio(struct dm_rq_target_io *tio, struct request *rq, + struct mapped_device *md) +{ + tio->md = md; + tio->ti = NULL; + tio->clone = NULL; + tio->orig = rq; + tio->error = 0; + /* + * Avoid initializing info for blk-mq; it passes + * target-specific data through info.ptr + * (see: dm_mq_init_request) + */ + if (!md->init_tio_pdu) + memset(&tio->info, 0, sizeof(tio->info)); + if (md->kworker_task) + init_kthread_work(&tio->work, map_tio_request); +} + +static struct dm_rq_target_io *dm_old_prep_tio(struct request *rq, + struct mapped_device *md, + gfp_t gfp_mask) +{ + struct dm_rq_target_io *tio; + int srcu_idx; + struct dm_table *table; + + tio = alloc_old_rq_tio(md, gfp_mask); + if (!tio) + return NULL; + + init_tio(tio, rq, md); + + table = dm_get_live_table(md, &srcu_idx); + /* + * Must clone a request if this .request_fn DM device + * is stacked on .request_fn device(s). + */ + if (!dm_table_all_blk_mq_devices(table)) { + if (!clone_old_rq(rq, md, tio, gfp_mask)) { + dm_put_live_table(md, srcu_idx); + free_old_rq_tio(tio); + return NULL; + } + } + dm_put_live_table(md, srcu_idx); + + return tio; +} + +/* + * Called with the queue lock held. + */ +static int dm_old_prep_fn(struct request_queue *q, struct request *rq) +{ + struct mapped_device *md = q->queuedata; + struct dm_rq_target_io *tio; + + if (unlikely(rq->special)) { + DMWARN("Already has something in rq->special."); + return BLKPREP_KILL; + } + + tio = dm_old_prep_tio(rq, md, GFP_ATOMIC); + if (!tio) + return BLKPREP_DEFER; + + rq->special = tio; + rq->cmd_flags |= REQ_DONTPREP; + + return BLKPREP_OK; +} + +/* + * Returns: + * 0 : the request has been processed + * DM_MAPIO_REQUEUE : the original request needs to be requeued + * < 0 : the request was completed due to failure + */ +static int map_request(struct dm_rq_target_io *tio, struct request *rq, + struct mapped_device *md) +{ + int r; + struct dm_target *ti = tio->ti; + struct request *clone = NULL; + + if (tio->clone) { + clone = tio->clone; + r = ti->type->map_rq(ti, clone, &tio->info); + } else { + r = ti->type->clone_and_map_rq(ti, rq, &tio->info, &clone); + if (r < 0) { + /* The target wants to complete the I/O */ + dm_kill_unmapped_request(rq, r); + return r; + } + if (r != DM_MAPIO_REMAPPED) + return r; + if (setup_clone(clone, rq, tio, GFP_ATOMIC)) { + /* -ENOMEM */ + ti->type->release_clone_rq(clone); + return DM_MAPIO_REQUEUE; + } + } + + switch (r) { + case DM_MAPIO_SUBMITTED: + /* The target has taken the I/O to submit by itself later */ + break; + case DM_MAPIO_REMAPPED: + /* The target has remapped the I/O so dispatch it */ + trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)), + blk_rq_pos(rq)); + dm_dispatch_clone_request(clone, rq); + break; + case DM_MAPIO_REQUEUE: + /* The target wants to requeue the I/O */ + dm_requeue_original_request(md, tio->orig); + break; + default: + if (r > 0) { + DMWARN("unimplemented target map return value: %d", r); + BUG(); + } + + /* The target wants to complete the I/O */ + dm_kill_unmapped_request(rq, r); + return r; + } + + return 0; +} + +static void dm_start_request(struct mapped_device *md, struct request *orig) +{ + if (!orig->q->mq_ops) + blk_start_request(orig); + else + blk_mq_start_request(orig); + atomic_inc(&md->pending[rq_data_dir(orig)]); + + if (md->seq_rq_merge_deadline_usecs) { + md->last_rq_pos = rq_end_sector(orig); + md->last_rq_rw = rq_data_dir(orig); + md->last_rq_start_time = ktime_get(); + } + + if (unlikely(dm_stats_used(&md->stats))) { + struct dm_rq_target_io *tio = tio_from_request(orig); + tio->duration_jiffies = jiffies; + tio->n_sectors = blk_rq_sectors(orig); + dm_stats_account_io(&md->stats, rq_data_dir(orig), + blk_rq_pos(orig), tio->n_sectors, false, 0, + &tio->stats_aux); + } + + /* + * Hold the md reference here for the in-flight I/O. + * We can't rely on the reference count by device opener, + * because the device may be closed during the request completion + * when all bios are completed. + * See the comment in rq_completed() too. + */ + dm_get(md); +} + +static void map_tio_request(struct kthread_work *work) +{ + struct dm_rq_target_io *tio = container_of(work, struct dm_rq_target_io, work); + struct request *rq = tio->orig; + struct mapped_device *md = tio->md; + + if (map_request(tio, rq, md) == DM_MAPIO_REQUEUE) + dm_requeue_original_request(md, rq); +} + +ssize_t dm_attr_rq_based_seq_io_merge_deadline_show(struct mapped_device *md, char *buf) +{ + return sprintf(buf, "%u\n", md->seq_rq_merge_deadline_usecs); +} + +#define MAX_SEQ_RQ_MERGE_DEADLINE_USECS 100000 + +ssize_t dm_attr_rq_based_seq_io_merge_deadline_store(struct mapped_device *md, + const char *buf, size_t count) +{ + unsigned deadline; + + if (dm_get_md_type(md) != DM_TYPE_REQUEST_BASED) + return count; + + if (kstrtouint(buf, 10, &deadline)) + return -EINVAL; + + if (deadline > MAX_SEQ_RQ_MERGE_DEADLINE_USECS) + deadline = MAX_SEQ_RQ_MERGE_DEADLINE_USECS; + + md->seq_rq_merge_deadline_usecs = deadline; + + return count; +} + +static bool dm_old_request_peeked_before_merge_deadline(struct mapped_device *md) +{ + ktime_t kt_deadline; + + if (!md->seq_rq_merge_deadline_usecs) + return false; + + kt_deadline = ns_to_ktime((u64)md->seq_rq_merge_deadline_usecs * NSEC_PER_USEC); + kt_deadline = ktime_add_safe(md->last_rq_start_time, kt_deadline); + + return !ktime_after(ktime_get(), kt_deadline); +} + +/* + * q->request_fn for old request-based dm. + * Called with the queue lock held. + */ +static void dm_old_request_fn(struct request_queue *q) +{ + struct mapped_device *md = q->queuedata; + struct dm_target *ti = md->immutable_target; + struct request *rq; + struct dm_rq_target_io *tio; + sector_t pos = 0; + + if (unlikely(!ti)) { + int srcu_idx; + struct dm_table *map = dm_get_live_table(md, &srcu_idx); + + ti = dm_table_find_target(map, pos); + dm_put_live_table(md, srcu_idx); + } + + /* + * For suspend, check blk_queue_stopped() and increment + * ->pending within a single queue_lock not to increment the + * number of in-flight I/Os after the queue is stopped in + * dm_suspend(). + */ + while (!blk_queue_stopped(q)) { + rq = blk_peek_request(q); + if (!rq) + return; + + /* always use block 0 to find the target for flushes for now */ + pos = 0; + if (req_op(rq) != REQ_OP_FLUSH) + pos = blk_rq_pos(rq); + + if ((dm_old_request_peeked_before_merge_deadline(md) && + md_in_flight(md) && rq->bio && rq->bio->bi_vcnt == 1 && + md->last_rq_pos == pos && md->last_rq_rw == rq_data_dir(rq)) || + (ti->type->busy && ti->type->busy(ti))) { + blk_delay_queue(q, 10); + return; + } + + dm_start_request(md, rq); + + tio = tio_from_request(rq); + /* Establish tio->ti before queuing work (map_tio_request) */ + tio->ti = ti; + queue_kthread_work(&md->kworker, &tio->work); + BUG_ON(!irqs_disabled()); + } +} + +/* + * Fully initialize a .request_fn request-based queue. + */ +int dm_old_init_request_queue(struct mapped_device *md) +{ + /* Fully initialize the queue */ + if (!blk_init_allocated_queue(md->queue, dm_old_request_fn, NULL)) + return -EINVAL; + + /* 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); + blk_queue_prep_rq(md->queue, dm_old_prep_fn); + + /* Initialize the request-based DM worker thread */ + init_kthread_worker(&md->kworker); + md->kworker_task = kthread_run(kthread_worker_fn, &md->kworker, + "kdmwork-%s", dm_device_name(md)); + if (IS_ERR(md->kworker_task)) + return PTR_ERR(md->kworker_task); + + elv_register_queue(md->queue); + + return 0; +} + +static int dm_mq_init_request(void *data, struct request *rq, + unsigned int hctx_idx, unsigned int request_idx, + unsigned int numa_node) +{ + struct mapped_device *md = data; + struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq); + + /* + * Must initialize md member of tio, otherwise it won't + * be available in dm_mq_queue_rq. + */ + tio->md = md; + + if (md->init_tio_pdu) { + /* target-specific per-io data is immediately after the tio */ + tio->info.ptr = tio + 1; + } + + return 0; +} + +static int dm_mq_queue_rq(struct blk_mq_hw_ctx *hctx, + const struct blk_mq_queue_data *bd) +{ + struct request *rq = bd->rq; + struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq); + struct mapped_device *md = tio->md; + struct dm_target *ti = md->immutable_target; + + if (unlikely(!ti)) { + int srcu_idx; + struct dm_table *map = dm_get_live_table(md, &srcu_idx); + + ti = dm_table_find_target(map, 0); + dm_put_live_table(md, srcu_idx); + } + + if (ti->type->busy && ti->type->busy(ti)) + return BLK_MQ_RQ_QUEUE_BUSY; + + dm_start_request(md, rq); + + /* Init tio using md established in .init_request */ + init_tio(tio, rq, md); + + /* + * Establish tio->ti before calling map_request(). + */ + tio->ti = ti; + + /* Direct call is fine since .queue_rq allows allocations */ + if (map_request(tio, rq, md) == DM_MAPIO_REQUEUE) { + /* Undo dm_start_request() before requeuing */ + rq_end_stats(md, rq); + rq_completed(md, rq_data_dir(rq), false); + return BLK_MQ_RQ_QUEUE_BUSY; + } + + return BLK_MQ_RQ_QUEUE_OK; +} + +static struct blk_mq_ops dm_mq_ops = { + .queue_rq = dm_mq_queue_rq, + .map_queue = blk_mq_map_queue, + .complete = dm_softirq_done, + .init_request = dm_mq_init_request, +}; + +int dm_mq_init_request_queue(struct mapped_device *md, struct dm_table *t) +{ + struct request_queue *q; + struct dm_target *immutable_tgt; + int err; + + if (!dm_table_all_blk_mq_devices(t)) { + DMERR("request-based dm-mq may only be stacked on blk-mq device(s)"); + return -EINVAL; + } + + md->tag_set = kzalloc_node(sizeof(struct blk_mq_tag_set), GFP_KERNEL, md->numa_node_id); + if (!md->tag_set) + return -ENOMEM; + + md->tag_set->ops = &dm_mq_ops; + md->tag_set->queue_depth = dm_get_blk_mq_queue_depth(); + md->tag_set->numa_node = md->numa_node_id; + md->tag_set->flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE; + md->tag_set->nr_hw_queues = dm_get_blk_mq_nr_hw_queues(); + md->tag_set->driver_data = md; + + md->tag_set->cmd_size = sizeof(struct dm_rq_target_io); + immutable_tgt = dm_table_get_immutable_target(t); + if (immutable_tgt && immutable_tgt->per_io_data_size) { + /* any target-specific per-io data is immediately after the tio */ + md->tag_set->cmd_size += immutable_tgt->per_io_data_size; + md->init_tio_pdu = true; + } + + err = blk_mq_alloc_tag_set(md->tag_set); + if (err) + goto out_kfree_tag_set; + + q = blk_mq_init_allocated_queue(md->tag_set, md->queue); + if (IS_ERR(q)) { + err = PTR_ERR(q); + goto out_tag_set; + } + dm_init_md_queue(md); + + /* backfill 'mq' sysfs registration normally done in blk_register_queue */ + blk_mq_register_disk(md->disk); + + return 0; + +out_tag_set: + blk_mq_free_tag_set(md->tag_set); +out_kfree_tag_set: + kfree(md->tag_set); + + return err; +} + +void dm_mq_cleanup_mapped_device(struct mapped_device *md) +{ + if (md->tag_set) { + blk_mq_free_tag_set(md->tag_set); + kfree(md->tag_set); + } +} + +module_param(reserved_rq_based_ios, uint, S_IRUGO | S_IWUSR); +MODULE_PARM_DESC(reserved_rq_based_ios, "Reserved IOs in request-based mempools"); + +module_param(use_blk_mq, bool, S_IRUGO | S_IWUSR); +MODULE_PARM_DESC(use_blk_mq, "Use block multiqueue for request-based DM devices"); + +module_param(dm_mq_nr_hw_queues, uint, S_IRUGO | S_IWUSR); +MODULE_PARM_DESC(dm_mq_nr_hw_queues, "Number of hardware queues for request-based dm-mq devices"); + +module_param(dm_mq_queue_depth, uint, S_IRUGO | S_IWUSR); +MODULE_PARM_DESC(dm_mq_queue_depth, "Queue depth for request-based dm-mq devices"); diff --git a/drivers/md/dm-rq.h b/drivers/md/dm-rq.h new file mode 100644 index 000000000000..9e6f0a3773d4 --- /dev/null +++ b/drivers/md/dm-rq.h @@ -0,0 +1,64 @@ +/* + * Internal header file for device mapper + * + * Copyright (C) 2016 Red Hat, Inc. All rights reserved. + * + * This file is released under the LGPL. + */ + +#ifndef DM_RQ_INTERNAL_H +#define DM_RQ_INTERNAL_H + +#include <linux/bio.h> +#include <linux/kthread.h> + +#include "dm-stats.h" + +struct mapped_device; + +/* + * One of these is allocated per request. + */ +struct dm_rq_target_io { + struct mapped_device *md; + struct dm_target *ti; + struct request *orig, *clone; + struct kthread_work work; + int error; + union map_info info; + struct dm_stats_aux stats_aux; + unsigned long duration_jiffies; + unsigned n_sectors; +}; + +/* + * For request-based dm - the bio clones we allocate are embedded in these + * structs. + * + * We allocate these with bio_alloc_bioset, using the front_pad parameter when + * the bioset is created - this means the bio has to come at the end of the + * struct. + */ +struct dm_rq_clone_bio_info { + struct bio *orig; + struct dm_rq_target_io *tio; + struct bio clone; +}; + +bool dm_use_blk_mq_default(void); +bool dm_use_blk_mq(struct mapped_device *md); + +int dm_old_init_request_queue(struct mapped_device *md); +int dm_mq_init_request_queue(struct mapped_device *md, struct dm_table *t); +void dm_mq_cleanup_mapped_device(struct mapped_device *md); + +void dm_start_queue(struct request_queue *q); +void dm_stop_queue(struct request_queue *q); + +unsigned dm_get_reserved_rq_based_ios(void); + +ssize_t dm_attr_rq_based_seq_io_merge_deadline_show(struct mapped_device *md, char *buf); +ssize_t dm_attr_rq_based_seq_io_merge_deadline_store(struct mapped_device *md, + const char *buf, size_t count); + +#endif diff --git a/drivers/md/dm-snap.c b/drivers/md/dm-snap.c index cc2f14b42ba4..731e1f5bd895 100644 --- a/drivers/md/dm-snap.c +++ b/drivers/md/dm-snap.c @@ -2302,6 +2302,13 @@ static int origin_map(struct dm_target *ti, struct bio *bio) return do_origin(o->dev, bio); } +static long origin_direct_access(struct dm_target *ti, sector_t sector, + void __pmem **kaddr, pfn_t *pfn, long size) +{ + DMWARN("device does not support dax."); + return -EIO; +} + /* * Set the target "max_io_len" field to the minimum of all the snapshots' * chunk sizes. @@ -2361,6 +2368,7 @@ static struct target_type origin_target = { .postsuspend = origin_postsuspend, .status = origin_status, .iterate_devices = origin_iterate_devices, + .direct_access = origin_direct_access, }; static struct target_type snapshot_target = { diff --git a/drivers/md/dm-stats.c b/drivers/md/dm-stats.c index 4fba26cd6bdb..38b05f23b96c 100644 --- a/drivers/md/dm-stats.c +++ b/drivers/md/dm-stats.c @@ -10,7 +10,7 @@ #include <linux/module.h> #include <linux/device-mapper.h> -#include "dm.h" +#include "dm-core.h" #include "dm-stats.h" #define DM_MSG_PREFIX "stats" diff --git a/drivers/md/dm-stripe.c b/drivers/md/dm-stripe.c index 48f1c01d7b9f..01bb9cf2a8c2 100644 --- a/drivers/md/dm-stripe.c +++ b/drivers/md/dm-stripe.c @@ -308,6 +308,29 @@ static int stripe_map(struct dm_target *ti, struct bio *bio) return DM_MAPIO_REMAPPED; } +static long stripe_direct_access(struct dm_target *ti, sector_t sector, + void __pmem **kaddr, pfn_t *pfn, long size) +{ + struct stripe_c *sc = ti->private; + uint32_t stripe; + struct block_device *bdev; + struct blk_dax_ctl dax = { + .size = size, + }; + long ret; + + stripe_map_sector(sc, sector, &stripe, &dax.sector); + + dax.sector += sc->stripe[stripe].physical_start; + bdev = sc->stripe[stripe].dev->bdev; + + ret = bdev_direct_access(bdev, &dax); + *kaddr = dax.addr; + *pfn = dax.pfn; + + return ret; +} + /* * Stripe status: * @@ -416,7 +439,7 @@ static void stripe_io_hints(struct dm_target *ti, static struct target_type stripe_target = { .name = "striped", - .version = {1, 5, 1}, + .version = {1, 6, 0}, .module = THIS_MODULE, .ctr = stripe_ctr, .dtr = stripe_dtr, @@ -425,6 +448,7 @@ static struct target_type stripe_target = { .status = stripe_status, .iterate_devices = stripe_iterate_devices, .io_hints = stripe_io_hints, + .direct_access = stripe_direct_access, }; int __init dm_stripe_init(void) diff --git a/drivers/md/dm-sysfs.c b/drivers/md/dm-sysfs.c index 7e818f5f1dc4..c209b8a19b84 100644 --- a/drivers/md/dm-sysfs.c +++ b/drivers/md/dm-sysfs.c @@ -6,7 +6,8 @@ #include <linux/sysfs.h> #include <linux/dm-ioctl.h> -#include "dm.h" +#include "dm-core.h" +#include "dm-rq.h" struct dm_sysfs_attr { struct attribute attr; diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c index 626a5ec04466..3e407a9cde1f 100644 --- a/drivers/md/dm-table.c +++ b/drivers/md/dm-table.c @@ -5,7 +5,7 @@ * This file is released under the GPL. */ -#include "dm.h" +#include "dm-core.h" #include <linux/module.h> #include <linux/vmalloc.h> @@ -43,8 +43,10 @@ struct dm_table { struct dm_target *targets; struct target_type *immutable_target_type; - unsigned integrity_supported:1; - unsigned singleton:1; + + bool integrity_supported:1; + bool singleton:1; + bool all_blk_mq:1; /* * Indicates the rw permissions for the new logical @@ -206,6 +208,7 @@ int dm_table_create(struct dm_table **result, fmode_t mode, return -ENOMEM; } + t->type = DM_TYPE_NONE; t->mode = mode; t->md = md; *result = t; @@ -703,7 +706,7 @@ int dm_table_add_target(struct dm_table *t, const char *type, dm_device_name(t->md), type); return -EINVAL; } - t->singleton = 1; + t->singleton = true; } if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) { @@ -824,22 +827,70 @@ void dm_consume_args(struct dm_arg_set *as, unsigned num_args) } EXPORT_SYMBOL(dm_consume_args); +static bool __table_type_bio_based(unsigned table_type) +{ + return (table_type == DM_TYPE_BIO_BASED || + table_type == DM_TYPE_DAX_BIO_BASED); +} + static bool __table_type_request_based(unsigned table_type) { return (table_type == DM_TYPE_REQUEST_BASED || table_type == DM_TYPE_MQ_REQUEST_BASED); } -static int dm_table_set_type(struct dm_table *t) +void dm_table_set_type(struct dm_table *t, unsigned type) +{ + t->type = type; +} +EXPORT_SYMBOL_GPL(dm_table_set_type); + +static int device_supports_dax(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); + + return q && blk_queue_dax(q); +} + +static bool dm_table_supports_dax(struct dm_table *t) +{ + struct dm_target *ti; + unsigned i = 0; + + /* Ensure that all targets support DAX. */ + while (i < dm_table_get_num_targets(t)) { + ti = dm_table_get_target(t, i++); + + if (!ti->type->direct_access) + return false; + + if (!ti->type->iterate_devices || + !ti->type->iterate_devices(ti, device_supports_dax, NULL)) + return false; + } + + return true; +} + +static int dm_table_determine_type(struct dm_table *t) { unsigned i; unsigned bio_based = 0, request_based = 0, hybrid = 0; - bool use_blk_mq = false; + bool verify_blk_mq = false; struct dm_target *tgt; struct dm_dev_internal *dd; - struct list_head *devices; + struct list_head *devices = dm_table_get_devices(t); unsigned live_md_type = dm_get_md_type(t->md); + if (t->type != DM_TYPE_NONE) { + /* target already set the table's type */ + if (t->type == DM_TYPE_BIO_BASED) + return 0; + BUG_ON(t->type == DM_TYPE_DAX_BIO_BASED); + goto verify_rq_based; + } + for (i = 0; i < t->num_targets; i++) { tgt = t->targets + i; if (dm_target_hybrid(tgt)) @@ -871,11 +922,27 @@ static int dm_table_set_type(struct dm_table *t) if (bio_based) { /* 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)) + t->type = DM_TYPE_DAX_BIO_BASED; return 0; } BUG_ON(!request_based); /* No targets in this table */ + if (list_empty(devices) && __table_type_request_based(live_md_type)) { + /* inherit live MD type */ + t->type = live_md_type; + return 0; + } + + /* + * The only way to establish DM_TYPE_MQ_REQUEST_BASED is by + * having a compatible target use dm_table_set_type. + */ + t->type = DM_TYPE_REQUEST_BASED; + +verify_rq_based: /* * Request-based dm supports only tables that have a single target now. * To support multiple targets, request splitting support is needed, @@ -888,7 +955,6 @@ static int dm_table_set_type(struct dm_table *t) } /* Non-request-stackable devices can't be used for request-based dm */ - devices = dm_table_get_devices(t); list_for_each_entry(dd, devices, list) { struct request_queue *q = bdev_get_queue(dd->dm_dev->bdev); @@ -899,10 +965,10 @@ static int dm_table_set_type(struct dm_table *t) } if (q->mq_ops) - use_blk_mq = true; + verify_blk_mq = true; } - if (use_blk_mq) { + if (verify_blk_mq) { /* verify _all_ devices in the table are blk-mq devices */ list_for_each_entry(dd, devices, list) if (!bdev_get_queue(dd->dm_dev->bdev)->mq_ops) { @@ -910,14 +976,9 @@ static int dm_table_set_type(struct dm_table *t) " are blk-mq request-stackable"); return -EINVAL; } - t->type = DM_TYPE_MQ_REQUEST_BASED; - } else if (list_empty(devices) && __table_type_request_based(live_md_type)) { - /* inherit live MD type */ - t->type = live_md_type; - - } else - t->type = DM_TYPE_REQUEST_BASED; + t->all_blk_mq = true; + } return 0; } @@ -956,14 +1017,19 @@ struct dm_target *dm_table_get_wildcard_target(struct dm_table *t) return NULL; } +bool dm_table_bio_based(struct dm_table *t) +{ + return __table_type_bio_based(dm_table_get_type(t)); +} + bool dm_table_request_based(struct dm_table *t) { return __table_type_request_based(dm_table_get_type(t)); } -bool dm_table_mq_request_based(struct dm_table *t) +bool dm_table_all_blk_mq_devices(struct dm_table *t) { - return dm_table_get_type(t) == DM_TYPE_MQ_REQUEST_BASED; + return t->all_blk_mq; } static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device *md) @@ -978,7 +1044,7 @@ static int dm_table_alloc_md_mempools(struct dm_table *t, struct mapped_device * return -EINVAL; } - if (type == DM_TYPE_BIO_BASED) + 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); @@ -1106,7 +1172,7 @@ static int dm_table_register_integrity(struct dm_table *t) return 0; if (!integrity_profile_exists(dm_disk(md))) { - t->integrity_supported = 1; + t->integrity_supported = true; /* * Register integrity profile during table load; we can do * this because the final profile must match during resume. @@ -1129,7 +1195,7 @@ static int dm_table_register_integrity(struct dm_table *t) } /* Preserve existing integrity profile */ - t->integrity_supported = 1; + t->integrity_supported = true; return 0; } @@ -1141,9 +1207,9 @@ int dm_table_complete(struct dm_table *t) { int r; - r = dm_table_set_type(t); + r = dm_table_determine_type(t); if (r) { - DMERR("unable to set table type"); + DMERR("unable to determine table type"); return r; } diff --git a/drivers/md/dm-target.c b/drivers/md/dm-target.c index a317dd884ba6..6eecd6b36f76 100644 --- a/drivers/md/dm-target.c +++ b/drivers/md/dm-target.c @@ -4,7 +4,7 @@ * This file is released under the GPL. */ -#include "dm.h" +#include "dm-core.h" #include <linux/module.h> #include <linux/init.h> @@ -148,9 +148,15 @@ static void io_err_release_clone_rq(struct request *clone) { } +static long io_err_direct_access(struct dm_target *ti, sector_t sector, + void __pmem **kaddr, pfn_t *pfn, long size) +{ + return -EIO; +} + static struct target_type error_target = { .name = "error", - .version = {1, 4, 0}, + .version = {1, 5, 0}, .features = DM_TARGET_WILDCARD, .ctr = io_err_ctr, .dtr = io_err_dtr, @@ -158,6 +164,7 @@ static struct target_type error_target = { .map_rq = io_err_map_rq, .clone_and_map_rq = io_err_clone_and_map_rq, .release_clone_rq = io_err_release_clone_rq, + .direct_access = io_err_direct_access, }; int __init dm_target_init(void) diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c index 43824d73366d..a15091a0d40c 100644 --- a/drivers/md/dm-thin-metadata.c +++ b/drivers/md/dm-thin-metadata.c @@ -1677,6 +1677,36 @@ int dm_pool_block_is_used(struct dm_pool_metadata *pmd, dm_block_t b, bool *resu return r; } +int dm_pool_inc_data_range(struct dm_pool_metadata *pmd, dm_block_t b, dm_block_t e) +{ + int r = 0; + + down_write(&pmd->root_lock); + for (; b != e; b++) { + r = dm_sm_inc_block(pmd->data_sm, b); + if (r) + break; + } + up_write(&pmd->root_lock); + + return r; +} + +int dm_pool_dec_data_range(struct dm_pool_metadata *pmd, dm_block_t b, dm_block_t e) +{ + int r = 0; + + down_write(&pmd->root_lock); + for (; b != e; b++) { + r = dm_sm_dec_block(pmd->data_sm, b); + if (r) + break; + } + up_write(&pmd->root_lock); + + return r; +} + bool dm_thin_changed_this_transaction(struct dm_thin_device *td) { int r; diff --git a/drivers/md/dm-thin-metadata.h b/drivers/md/dm-thin-metadata.h index a938babe4258..35e954ea20a9 100644 --- a/drivers/md/dm-thin-metadata.h +++ b/drivers/md/dm-thin-metadata.h @@ -197,6 +197,9 @@ int dm_pool_get_data_dev_size(struct dm_pool_metadata *pmd, dm_block_t *result); int dm_pool_block_is_used(struct dm_pool_metadata *pmd, dm_block_t b, bool *result); +int dm_pool_inc_data_range(struct dm_pool_metadata *pmd, dm_block_t b, dm_block_t e); +int dm_pool_dec_data_range(struct dm_pool_metadata *pmd, dm_block_t b, dm_block_t e); + /* * Returns -ENOSPC if the new size is too small and already allocated * blocks would be lost. diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c index 5f9e3d799d66..197ea2003400 100644 --- a/drivers/md/dm-thin.c +++ b/drivers/md/dm-thin.c @@ -253,6 +253,7 @@ struct pool { struct bio_list deferred_flush_bios; struct list_head prepared_mappings; struct list_head prepared_discards; + struct list_head prepared_discards_pt2; struct list_head active_thins; struct dm_deferred_set *shared_read_ds; @@ -269,6 +270,7 @@ struct pool { process_mapping_fn process_prepared_mapping; process_mapping_fn process_prepared_discard; + process_mapping_fn process_prepared_discard_pt2; struct dm_bio_prison_cell **cell_sort_array; }; @@ -1001,7 +1003,8 @@ static void process_prepared_discard_no_passdown(struct dm_thin_new_mapping *m) /*----------------------------------------------------------------*/ -static void passdown_double_checking_shared_status(struct dm_thin_new_mapping *m) +static void passdown_double_checking_shared_status(struct dm_thin_new_mapping *m, + struct bio *discard_parent) { /* * We've already unmapped this range of blocks, but before we @@ -1014,7 +1017,7 @@ static void passdown_double_checking_shared_status(struct dm_thin_new_mapping *m dm_block_t b = m->data_block, e, end = m->data_block + m->virt_end - m->virt_begin; struct discard_op op; - begin_discard(&op, tc, m->bio); + begin_discard(&op, tc, discard_parent); while (b != end) { /* find start of unmapped run */ for (; b < end; b++) { @@ -1049,28 +1052,101 @@ out: end_discard(&op, r); } -static void process_prepared_discard_passdown(struct dm_thin_new_mapping *m) +static void queue_passdown_pt2(struct dm_thin_new_mapping *m) +{ + unsigned long flags; + struct pool *pool = m->tc->pool; + + spin_lock_irqsave(&pool->lock, flags); + list_add_tail(&m->list, &pool->prepared_discards_pt2); + spin_unlock_irqrestore(&pool->lock, flags); + wake_worker(pool); +} + +static void passdown_endio(struct bio *bio) +{ + /* + * It doesn't matter if the passdown discard failed, we still want + * to unmap (we ignore err). + */ + queue_passdown_pt2(bio->bi_private); +} + +static void process_prepared_discard_passdown_pt1(struct dm_thin_new_mapping *m) { int r; struct thin_c *tc = m->tc; struct pool *pool = tc->pool; + struct bio *discard_parent; + dm_block_t data_end = m->data_block + (m->virt_end - m->virt_begin); + /* + * Only this thread allocates blocks, so we can be sure that the + * newly unmapped blocks will not be allocated before the end of + * the function. + */ r = dm_thin_remove_range(tc->td, m->virt_begin, m->virt_end); if (r) { metadata_operation_failed(pool, "dm_thin_remove_range", r); bio_io_error(m->bio); + cell_defer_no_holder(tc, m->cell); + mempool_free(m, pool->mapping_pool); + return; + } - } else if (m->maybe_shared) { - passdown_double_checking_shared_status(m); + discard_parent = bio_alloc(GFP_NOIO, 1); + if (!discard_parent) { + DMWARN("%s: unable to allocate top level discard bio for passdown. Skipping passdown.", + dm_device_name(tc->pool->pool_md)); + queue_passdown_pt2(m); } else { - struct discard_op op; - begin_discard(&op, tc, m->bio); - r = issue_discard(&op, m->data_block, - m->data_block + (m->virt_end - m->virt_begin)); - end_discard(&op, r); + discard_parent->bi_end_io = passdown_endio; + discard_parent->bi_private = m; + + if (m->maybe_shared) + passdown_double_checking_shared_status(m, discard_parent); + else { + struct discard_op op; + + begin_discard(&op, tc, discard_parent); + r = issue_discard(&op, m->data_block, data_end); + end_discard(&op, r); + } } + /* + * Increment the unmapped blocks. This prevents a race between the + * passdown io and reallocation of freed blocks. + */ + r = dm_pool_inc_data_range(pool->pmd, m->data_block, data_end); + if (r) { + metadata_operation_failed(pool, "dm_pool_inc_data_range", r); + bio_io_error(m->bio); + cell_defer_no_holder(tc, m->cell); + mempool_free(m, pool->mapping_pool); + return; + } +} + +static void process_prepared_discard_passdown_pt2(struct dm_thin_new_mapping *m) +{ + int r; + struct thin_c *tc = m->tc; + struct pool *pool = tc->pool; + + /* + * The passdown has completed, so now we can decrement all those + * unmapped blocks. + */ + r = dm_pool_dec_data_range(pool->pmd, m->data_block, + m->data_block + (m->virt_end - m->virt_begin)); + if (r) { + metadata_operation_failed(pool, "dm_pool_dec_data_range", r); + bio_io_error(m->bio); + } else + bio_endio(m->bio); + cell_defer_no_holder(tc, m->cell); mempool_free(m, pool->mapping_pool); } @@ -2215,6 +2291,8 @@ static void do_worker(struct work_struct *ws) throttle_work_update(&pool->throttle); process_prepared(pool, &pool->prepared_discards, &pool->process_prepared_discard); throttle_work_update(&pool->throttle); + process_prepared(pool, &pool->prepared_discards_pt2, &pool->process_prepared_discard_pt2); + throttle_work_update(&pool->throttle); process_deferred_bios(pool); throttle_work_complete(&pool->throttle); } @@ -2343,7 +2421,8 @@ static void set_discard_callbacks(struct pool *pool) if (passdown_enabled(pt)) { pool->process_discard_cell = process_discard_cell_passdown; - pool->process_prepared_discard = process_prepared_discard_passdown; + pool->process_prepared_discard = process_prepared_discard_passdown_pt1; + pool->process_prepared_discard_pt2 = process_prepared_discard_passdown_pt2; } else { pool->process_discard_cell = process_discard_cell_no_passdown; pool->process_prepared_discard = process_prepared_discard_no_passdown; @@ -2830,6 +2909,7 @@ static struct pool *pool_create(struct mapped_device *pool_md, bio_list_init(&pool->deferred_flush_bios); INIT_LIST_HEAD(&pool->prepared_mappings); INIT_LIST_HEAD(&pool->prepared_discards); + INIT_LIST_HEAD(&pool->prepared_discards_pt2); INIT_LIST_HEAD(&pool->active_thins); pool->low_water_triggered = false; pool->suspended = true; diff --git a/drivers/md/dm-verity-fec.c b/drivers/md/dm-verity-fec.c index 459a9f8905ed..0f0eb8a3d922 100644 --- a/drivers/md/dm-verity-fec.c +++ b/drivers/md/dm-verity-fec.c @@ -453,9 +453,7 @@ int verity_fec_decode(struct dm_verity *v, struct dm_verity_io *io, */ offset = block << v->data_dev_block_bits; - - res = offset; - div64_u64(res, v->fec->rounds << v->data_dev_block_bits); + res = div64_u64(offset, v->fec->rounds << v->data_dev_block_bits); /* * The base RS block we can feed to the interleaver to find out all diff --git a/drivers/md/dm.c b/drivers/md/dm.c index 812fd5984eea..ceb69fc0b10b 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -5,13 +5,13 @@ * This file is released under the GPL. */ -#include "dm.h" +#include "dm-core.h" +#include "dm-rq.h" #include "dm-uevent.h" #include <linux/init.h> #include <linux/module.h> #include <linux/mutex.h> -#include <linux/moduleparam.h> #include <linux/blkpg.h> #include <linux/bio.h> #include <linux/mempool.h> @@ -20,14 +20,8 @@ #include <linux/hdreg.h> #include <linux/delay.h> #include <linux/wait.h> -#include <linux/kthread.h> -#include <linux/ktime.h> -#include <linux/elevator.h> /* for rq_end_sector() */ -#include <linux/blk-mq.h> #include <linux/pr.h> -#include <trace/events/block.h> - #define DM_MSG_PREFIX "core" #ifdef CONFIG_PRINTK @@ -63,7 +57,6 @@ static DECLARE_WORK(deferred_remove_work, do_deferred_remove); static struct workqueue_struct *deferred_remove_workqueue; /* - * For bio-based dm. * One of these is allocated per bio. */ struct dm_io { @@ -76,36 +69,6 @@ struct dm_io { struct dm_stats_aux stats_aux; }; -/* - * For request-based dm. - * One of these is allocated per request. - */ -struct dm_rq_target_io { - struct mapped_device *md; - struct dm_target *ti; - struct request *orig, *clone; - struct kthread_work work; - int error; - union map_info info; - struct dm_stats_aux stats_aux; - unsigned long duration_jiffies; - unsigned n_sectors; -}; - -/* - * For request-based dm - the bio clones we allocate are embedded in these - * structs. - * - * We allocate these with bio_alloc_bioset, using the front_pad parameter when - * the bioset is created - this means the bio has to come at the end of the - * struct. - */ -struct dm_rq_clone_bio_info { - struct bio *orig; - struct dm_rq_target_io *tio; - struct bio clone; -}; - #define MINOR_ALLOCED ((void *)-1) /* @@ -120,130 +83,9 @@ struct dm_rq_clone_bio_info { #define DMF_DEFERRED_REMOVE 6 #define DMF_SUSPENDED_INTERNALLY 7 -/* - * Work processed by per-device workqueue. - */ -struct mapped_device { - struct srcu_struct io_barrier; - struct mutex suspend_lock; - - /* - * The current mapping (struct dm_table *). - * Use dm_get_live_table{_fast} or take suspend_lock for - * dereference. - */ - void __rcu *map; - - struct list_head table_devices; - struct mutex table_devices_lock; - - unsigned long flags; - - struct request_queue *queue; - int numa_node_id; - - unsigned type; - /* Protect queue and type against concurrent access. */ - struct mutex type_lock; - - atomic_t holders; - atomic_t open_count; - - struct dm_target *immutable_target; - struct target_type *immutable_target_type; - - struct gendisk *disk; - char name[16]; - - void *interface_ptr; - - /* - * A list of ios that arrived while we were suspended. - */ - atomic_t pending[2]; - wait_queue_head_t wait; - struct work_struct work; - spinlock_t deferred_lock; - struct bio_list deferred; - - /* - * Event handling. - */ - wait_queue_head_t eventq; - atomic_t event_nr; - atomic_t uevent_seq; - struct list_head uevent_list; - spinlock_t uevent_lock; /* Protect access to uevent_list */ - - /* the number of internal suspends */ - unsigned internal_suspend_count; - - /* - * Processing queue (flush) - */ - struct workqueue_struct *wq; - - /* - * io objects are allocated from here. - */ - mempool_t *io_pool; - mempool_t *rq_pool; - - struct bio_set *bs; - - /* - * freeze/thaw support require holding onto a super block - */ - struct super_block *frozen_sb; - - /* forced geometry settings */ - struct hd_geometry geometry; - - struct block_device *bdev; - - /* kobject and completion */ - struct dm_kobject_holder kobj_holder; - - /* zero-length flush that will be cloned and submitted to targets */ - struct bio flush_bio; - - struct dm_stats stats; - - struct kthread_worker kworker; - struct task_struct *kworker_task; - - /* for request-based merge heuristic in dm_request_fn() */ - unsigned seq_rq_merge_deadline_usecs; - int last_rq_rw; - sector_t last_rq_pos; - ktime_t last_rq_start_time; - - /* for blk-mq request-based DM support */ - struct blk_mq_tag_set *tag_set; - bool use_blk_mq:1; - bool init_tio_pdu:1; -}; - -#ifdef CONFIG_DM_MQ_DEFAULT -static bool use_blk_mq = true; -#else -static bool use_blk_mq = false; -#endif - -#define DM_MQ_NR_HW_QUEUES 1 -#define DM_MQ_QUEUE_DEPTH 2048 #define DM_NUMA_NODE NUMA_NO_NODE - -static unsigned dm_mq_nr_hw_queues = DM_MQ_NR_HW_QUEUES; -static unsigned dm_mq_queue_depth = DM_MQ_QUEUE_DEPTH; static int dm_numa_node = DM_NUMA_NODE; -bool dm_use_blk_mq(struct mapped_device *md) -{ - return md->use_blk_mq; -} -EXPORT_SYMBOL_GPL(dm_use_blk_mq); - /* * For mempools pre-allocation at the table loading time. */ @@ -259,9 +101,6 @@ struct table_device { struct dm_dev dm_dev; }; -#define RESERVED_BIO_BASED_IOS 16 -#define RESERVED_REQUEST_BASED_IOS 256 -#define RESERVED_MAX_IOS 1024 static struct kmem_cache *_io_cache; static struct kmem_cache *_rq_tio_cache; static struct kmem_cache *_rq_cache; @@ -269,13 +108,9 @@ static struct kmem_cache *_rq_cache; /* * Bio-based DM's mempools' reserved IOs set by the user. */ +#define RESERVED_BIO_BASED_IOS 16 static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS; -/* - * Request-based DM's mempools' reserved IOs set by the user. - */ -static unsigned reserved_rq_based_ios = RESERVED_REQUEST_BASED_IOS; - static int __dm_get_module_param_int(int *module_param, int min, int max) { int param = ACCESS_ONCE(*module_param); @@ -297,8 +132,8 @@ static int __dm_get_module_param_int(int *module_param, int min, int max) return param; } -static unsigned __dm_get_module_param(unsigned *module_param, - unsigned def, unsigned max) +unsigned __dm_get_module_param(unsigned *module_param, + unsigned def, unsigned max) { unsigned param = ACCESS_ONCE(*module_param); unsigned modified_param = 0; @@ -319,28 +154,10 @@ static unsigned __dm_get_module_param(unsigned *module_param, unsigned dm_get_reserved_bio_based_ios(void) { return __dm_get_module_param(&reserved_bio_based_ios, - RESERVED_BIO_BASED_IOS, RESERVED_MAX_IOS); + RESERVED_BIO_BASED_IOS, DM_RESERVED_MAX_IOS); } EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios); -unsigned dm_get_reserved_rq_based_ios(void) -{ - return __dm_get_module_param(&reserved_rq_based_ios, - RESERVED_REQUEST_BASED_IOS, RESERVED_MAX_IOS); -} -EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios); - -static unsigned dm_get_blk_mq_nr_hw_queues(void) -{ - return __dm_get_module_param(&dm_mq_nr_hw_queues, 1, 32); -} - -static unsigned dm_get_blk_mq_queue_depth(void) -{ - return __dm_get_module_param(&dm_mq_queue_depth, - DM_MQ_QUEUE_DEPTH, BLK_MQ_MAX_DEPTH); -} - static unsigned dm_get_numa_node(void) { return __dm_get_module_param_int(&dm_numa_node, @@ -679,29 +496,7 @@ static void free_tio(struct dm_target_io *tio) bio_put(&tio->clone); } -static struct dm_rq_target_io *alloc_old_rq_tio(struct mapped_device *md, - gfp_t gfp_mask) -{ - return mempool_alloc(md->io_pool, gfp_mask); -} - -static void free_old_rq_tio(struct dm_rq_target_io *tio) -{ - mempool_free(tio, tio->md->io_pool); -} - -static struct request *alloc_old_clone_request(struct mapped_device *md, - gfp_t gfp_mask) -{ - return mempool_alloc(md->rq_pool, gfp_mask); -} - -static void free_old_clone_request(struct mapped_device *md, struct request *rq) -{ - mempool_free(rq, md->rq_pool); -} - -static int md_in_flight(struct mapped_device *md) +int md_in_flight(struct mapped_device *md) { return atomic_read(&md->pending[READ]) + atomic_read(&md->pending[WRITE]); @@ -1019,7 +814,7 @@ static void dec_pending(struct dm_io *io, int error) } } -static void disable_write_same(struct mapped_device *md) +void disable_write_same(struct mapped_device *md) { struct queue_limits *limits = dm_get_queue_limits(md); @@ -1062,371 +857,6 @@ static void clone_endio(struct bio *bio) } /* - * Partial completion handling for request-based dm - */ -static void end_clone_bio(struct bio *clone) -{ - struct dm_rq_clone_bio_info *info = - container_of(clone, struct dm_rq_clone_bio_info, clone); - struct dm_rq_target_io *tio = info->tio; - struct bio *bio = info->orig; - unsigned int nr_bytes = info->orig->bi_iter.bi_size; - int error = clone->bi_error; - - bio_put(clone); - - if (tio->error) - /* - * An error has already been detected on the request. - * Once error occurred, just let clone->end_io() handle - * the remainder. - */ - return; - else if (error) { - /* - * Don't notice the error to the upper layer yet. - * The error handling decision is made by the target driver, - * when the request is completed. - */ - tio->error = error; - return; - } - - /* - * I/O for the bio successfully completed. - * Notice the data completion to the upper layer. - */ - - /* - * bios are processed from the head of the list. - * So the completing bio should always be rq->bio. - * If it's not, something wrong is happening. - */ - if (tio->orig->bio != bio) - DMERR("bio completion is going in the middle of the request"); - - /* - * Update the original request. - * Do not use blk_end_request() here, because it may complete - * the original request before the clone, and break the ordering. - */ - blk_update_request(tio->orig, 0, nr_bytes); -} - -static struct dm_rq_target_io *tio_from_request(struct request *rq) -{ - return (rq->q->mq_ops ? blk_mq_rq_to_pdu(rq) : rq->special); -} - -static void rq_end_stats(struct mapped_device *md, struct request *orig) -{ - if (unlikely(dm_stats_used(&md->stats))) { - struct dm_rq_target_io *tio = tio_from_request(orig); - tio->duration_jiffies = jiffies - tio->duration_jiffies; - dm_stats_account_io(&md->stats, rq_data_dir(orig), - blk_rq_pos(orig), tio->n_sectors, true, - tio->duration_jiffies, &tio->stats_aux); - } -} - -/* - * Don't touch any member of the md after calling this function because - * the md may be freed in dm_put() at the end of this function. - * Or do dm_get() before calling this function and dm_put() later. - */ -static void rq_completed(struct mapped_device *md, int rw, bool run_queue) -{ - atomic_dec(&md->pending[rw]); - - /* nudge anyone waiting on suspend queue */ - if (!md_in_flight(md)) - wake_up(&md->wait); - - /* - * Run this off this callpath, as drivers could invoke end_io while - * inside their request_fn (and holding the queue lock). Calling - * back into ->request_fn() could deadlock attempting to grab the - * queue lock again. - */ - if (!md->queue->mq_ops && run_queue) - blk_run_queue_async(md->queue); - - /* - * dm_put() must be at the end of this function. See the comment above - */ - dm_put(md); -} - -static void free_rq_clone(struct request *clone) -{ - struct dm_rq_target_io *tio = clone->end_io_data; - struct mapped_device *md = tio->md; - - blk_rq_unprep_clone(clone); - - if (md->type == DM_TYPE_MQ_REQUEST_BASED) - /* stacked on blk-mq queue(s) */ - tio->ti->type->release_clone_rq(clone); - else if (!md->queue->mq_ops) - /* request_fn queue stacked on request_fn queue(s) */ - free_old_clone_request(md, clone); - - if (!md->queue->mq_ops) - free_old_rq_tio(tio); -} - -/* - * Complete the clone and the original request. - * Must be called without clone's queue lock held, - * see end_clone_request() for more details. - */ -static void dm_end_request(struct request *clone, int error) -{ - int rw = rq_data_dir(clone); - struct dm_rq_target_io *tio = clone->end_io_data; - struct mapped_device *md = tio->md; - struct request *rq = tio->orig; - - if (rq->cmd_type == REQ_TYPE_BLOCK_PC) { - rq->errors = clone->errors; - rq->resid_len = clone->resid_len; - - if (rq->sense) - /* - * We are using the sense buffer of the original - * request. - * So setting the length of the sense data is enough. - */ - rq->sense_len = clone->sense_len; - } - - free_rq_clone(clone); - rq_end_stats(md, rq); - if (!rq->q->mq_ops) - blk_end_request_all(rq, error); - else - blk_mq_end_request(rq, error); - rq_completed(md, rw, true); -} - -static void dm_unprep_request(struct request *rq) -{ - struct dm_rq_target_io *tio = tio_from_request(rq); - struct request *clone = tio->clone; - - if (!rq->q->mq_ops) { - rq->special = NULL; - rq->cmd_flags &= ~REQ_DONTPREP; - } - - if (clone) - free_rq_clone(clone); - else if (!tio->md->queue->mq_ops) - free_old_rq_tio(tio); -} - -/* - * Requeue the original request of a clone. - */ -static void dm_old_requeue_request(struct request *rq) -{ - struct request_queue *q = rq->q; - unsigned long flags; - - spin_lock_irqsave(q->queue_lock, flags); - blk_requeue_request(q, rq); - blk_run_queue_async(q); - spin_unlock_irqrestore(q->queue_lock, flags); -} - -static void dm_mq_requeue_request(struct request *rq) -{ - struct request_queue *q = rq->q; - unsigned long flags; - - blk_mq_requeue_request(rq); - spin_lock_irqsave(q->queue_lock, flags); - if (!blk_queue_stopped(q)) - blk_mq_kick_requeue_list(q); - spin_unlock_irqrestore(q->queue_lock, flags); -} - -static void dm_requeue_original_request(struct mapped_device *md, - struct request *rq) -{ - int rw = rq_data_dir(rq); - - rq_end_stats(md, rq); - dm_unprep_request(rq); - - if (!rq->q->mq_ops) - dm_old_requeue_request(rq); - else - dm_mq_requeue_request(rq); - - rq_completed(md, rw, false); -} - -static void dm_old_stop_queue(struct request_queue *q) -{ - unsigned long flags; - - spin_lock_irqsave(q->queue_lock, flags); - if (blk_queue_stopped(q)) { - spin_unlock_irqrestore(q->queue_lock, flags); - return; - } - - blk_stop_queue(q); - spin_unlock_irqrestore(q->queue_lock, flags); -} - -static void dm_stop_queue(struct request_queue *q) -{ - if (!q->mq_ops) - dm_old_stop_queue(q); - else - blk_mq_stop_hw_queues(q); -} - -static void dm_old_start_queue(struct request_queue *q) -{ - unsigned long flags; - - spin_lock_irqsave(q->queue_lock, flags); - if (blk_queue_stopped(q)) - blk_start_queue(q); - spin_unlock_irqrestore(q->queue_lock, flags); -} - -static void dm_start_queue(struct request_queue *q) -{ - if (!q->mq_ops) - dm_old_start_queue(q); - else { - blk_mq_start_stopped_hw_queues(q, true); - blk_mq_kick_requeue_list(q); - } -} - -static void dm_done(struct request *clone, int error, bool mapped) -{ - int r = error; - struct dm_rq_target_io *tio = clone->end_io_data; - dm_request_endio_fn rq_end_io = NULL; - - if (tio->ti) { - rq_end_io = tio->ti->type->rq_end_io; - - if (mapped && rq_end_io) - r = rq_end_io(tio->ti, clone, error, &tio->info); - } - - if (unlikely(r == -EREMOTEIO && (req_op(clone) == REQ_OP_WRITE_SAME) && - !clone->q->limits.max_write_same_sectors)) - disable_write_same(tio->md); - - if (r <= 0) - /* The target wants to complete the I/O */ - dm_end_request(clone, r); - else if (r == DM_ENDIO_INCOMPLETE) - /* The target will handle the I/O */ - return; - else if (r == DM_ENDIO_REQUEUE) - /* The target wants to requeue the I/O */ - dm_requeue_original_request(tio->md, tio->orig); - else { - DMWARN("unimplemented target endio return value: %d", r); - BUG(); - } -} - -/* - * Request completion handler for request-based dm - */ -static void dm_softirq_done(struct request *rq) -{ - bool mapped = true; - struct dm_rq_target_io *tio = tio_from_request(rq); - struct request *clone = tio->clone; - int rw; - - if (!clone) { - rq_end_stats(tio->md, rq); - rw = rq_data_dir(rq); - if (!rq->q->mq_ops) { - blk_end_request_all(rq, tio->error); - rq_completed(tio->md, rw, false); - free_old_rq_tio(tio); - } else { - blk_mq_end_request(rq, tio->error); - rq_completed(tio->md, rw, false); - } - return; - } - - if (rq->cmd_flags & REQ_FAILED) - mapped = false; - - dm_done(clone, tio->error, mapped); -} - -/* - * Complete the clone and the original request with the error status - * through softirq context. - */ -static void dm_complete_request(struct request *rq, int error) -{ - struct dm_rq_target_io *tio = tio_from_request(rq); - - tio->error = error; - if (!rq->q->mq_ops) - blk_complete_request(rq); - else - blk_mq_complete_request(rq, error); -} - -/* - * Complete the not-mapped clone and the original request with the error status - * through softirq context. - * Target's rq_end_io() function isn't called. - * This may be used when the target's map_rq() or clone_and_map_rq() functions fail. - */ -static void dm_kill_unmapped_request(struct request *rq, int error) -{ - rq->cmd_flags |= REQ_FAILED; - dm_complete_request(rq, error); -} - -/* - * Called with the clone's queue lock held (in the case of .request_fn) - */ -static void end_clone_request(struct request *clone, int error) -{ - struct dm_rq_target_io *tio = clone->end_io_data; - - if (!clone->q->mq_ops) { - /* - * For just cleaning up the information of the queue in which - * the clone was dispatched. - * The clone is *NOT* freed actually here because it is alloced - * from dm own mempool (REQ_ALLOCED isn't set). - */ - __blk_put_request(clone->q, clone); - } - - /* - * Actual request completion is done in a softirq context which doesn't - * hold the clone's queue lock. Otherwise, deadlock could occur because: - * - another request may be submitted by the upper level driver - * of the stacking during the completion - * - the submission which requires queue lock may be done - * against this clone's queue - */ - dm_complete_request(tio->orig, error); -} - -/* * Return maximum size of I/O possible at the supplied sector up to the current * target boundary. */ @@ -1475,6 +905,33 @@ int dm_set_target_max_io_len(struct dm_target *ti, sector_t len) } EXPORT_SYMBOL_GPL(dm_set_target_max_io_len); +static long dm_blk_direct_access(struct block_device *bdev, sector_t sector, + void __pmem **kaddr, pfn_t *pfn, long size) +{ + struct mapped_device *md = bdev->bd_disk->private_data; + struct dm_table *map; + struct dm_target *ti; + int srcu_idx; + long len, ret = -EIO; + + map = dm_get_live_table(md, &srcu_idx); + if (!map) + goto out; + + ti = dm_table_find_target(map, sector); + if (!dm_target_is_valid(ti)) + goto out; + + len = max_io_len(sector, ti) << SECTOR_SHIFT; + size = min(len, size); + + if (ti->type->direct_access) + ret = ti->type->direct_access(ti, sector, kaddr, pfn, size); +out: + dm_put_live_table(md, srcu_idx); + return min(ret, size); +} + /* * A target may call dm_accept_partial_bio only from the map routine. It is * allowed for all bio types except REQ_PREFLUSH. @@ -1845,353 +1302,6 @@ static blk_qc_t dm_make_request(struct request_queue *q, struct bio *bio) return BLK_QC_T_NONE; } -int dm_request_based(struct mapped_device *md) -{ - return blk_queue_stackable(md->queue); -} - -static void dm_dispatch_clone_request(struct request *clone, struct request *rq) -{ - int r; - - if (blk_queue_io_stat(clone->q)) - clone->cmd_flags |= REQ_IO_STAT; - - clone->start_time = jiffies; - r = blk_insert_cloned_request(clone->q, clone); - if (r) - /* must complete clone in terms of original request */ - dm_complete_request(rq, r); -} - -static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig, - void *data) -{ - struct dm_rq_target_io *tio = data; - struct dm_rq_clone_bio_info *info = - container_of(bio, struct dm_rq_clone_bio_info, clone); - - info->orig = bio_orig; - info->tio = tio; - bio->bi_end_io = end_clone_bio; - - return 0; -} - -static int setup_clone(struct request *clone, struct request *rq, - struct dm_rq_target_io *tio, gfp_t gfp_mask) -{ - int r; - - r = blk_rq_prep_clone(clone, rq, tio->md->bs, gfp_mask, - dm_rq_bio_constructor, tio); - if (r) - return r; - - clone->cmd = rq->cmd; - clone->cmd_len = rq->cmd_len; - clone->sense = rq->sense; - clone->end_io = end_clone_request; - clone->end_io_data = tio; - - tio->clone = clone; - - return 0; -} - -static struct request *clone_old_rq(struct request *rq, struct mapped_device *md, - struct dm_rq_target_io *tio, gfp_t gfp_mask) -{ - /* - * Create clone for use with .request_fn request_queue - */ - struct request *clone; - - clone = alloc_old_clone_request(md, gfp_mask); - if (!clone) - return NULL; - - blk_rq_init(NULL, clone); - if (setup_clone(clone, rq, tio, gfp_mask)) { - /* -ENOMEM */ - free_old_clone_request(md, clone); - return NULL; - } - - return clone; -} - -static void map_tio_request(struct kthread_work *work); - -static void init_tio(struct dm_rq_target_io *tio, struct request *rq, - struct mapped_device *md) -{ - tio->md = md; - tio->ti = NULL; - tio->clone = NULL; - tio->orig = rq; - tio->error = 0; - /* - * Avoid initializing info for blk-mq; it passes - * target-specific data through info.ptr - * (see: dm_mq_init_request) - */ - if (!md->init_tio_pdu) - memset(&tio->info, 0, sizeof(tio->info)); - if (md->kworker_task) - init_kthread_work(&tio->work, map_tio_request); -} - -static struct dm_rq_target_io *dm_old_prep_tio(struct request *rq, - struct mapped_device *md, - gfp_t gfp_mask) -{ - struct dm_rq_target_io *tio; - int srcu_idx; - struct dm_table *table; - - tio = alloc_old_rq_tio(md, gfp_mask); - if (!tio) - return NULL; - - init_tio(tio, rq, md); - - table = dm_get_live_table(md, &srcu_idx); - /* - * Must clone a request if this .request_fn DM device - * is stacked on .request_fn device(s). - */ - if (!dm_table_mq_request_based(table)) { - if (!clone_old_rq(rq, md, tio, gfp_mask)) { - dm_put_live_table(md, srcu_idx); - free_old_rq_tio(tio); - return NULL; - } - } - dm_put_live_table(md, srcu_idx); - - return tio; -} - -/* - * Called with the queue lock held. - */ -static int dm_old_prep_fn(struct request_queue *q, struct request *rq) -{ - struct mapped_device *md = q->queuedata; - struct dm_rq_target_io *tio; - - if (unlikely(rq->special)) { - DMWARN("Already has something in rq->special."); - return BLKPREP_KILL; - } - - tio = dm_old_prep_tio(rq, md, GFP_ATOMIC); - if (!tio) - return BLKPREP_DEFER; - - rq->special = tio; - rq->cmd_flags |= REQ_DONTPREP; - - return BLKPREP_OK; -} - -/* - * Returns: - * 0 : the request has been processed - * DM_MAPIO_REQUEUE : the original request needs to be requeued - * < 0 : the request was completed due to failure - */ -static int map_request(struct dm_rq_target_io *tio, struct request *rq, - struct mapped_device *md) -{ - int r; - struct dm_target *ti = tio->ti; - struct request *clone = NULL; - - if (tio->clone) { - clone = tio->clone; - r = ti->type->map_rq(ti, clone, &tio->info); - } else { - r = ti->type->clone_and_map_rq(ti, rq, &tio->info, &clone); - if (r < 0) { - /* The target wants to complete the I/O */ - dm_kill_unmapped_request(rq, r); - return r; - } - if (r != DM_MAPIO_REMAPPED) - return r; - if (setup_clone(clone, rq, tio, GFP_ATOMIC)) { - /* -ENOMEM */ - ti->type->release_clone_rq(clone); - return DM_MAPIO_REQUEUE; - } - } - - switch (r) { - case DM_MAPIO_SUBMITTED: - /* The target has taken the I/O to submit by itself later */ - break; - case DM_MAPIO_REMAPPED: - /* The target has remapped the I/O so dispatch it */ - trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)), - blk_rq_pos(rq)); - dm_dispatch_clone_request(clone, rq); - break; - case DM_MAPIO_REQUEUE: - /* The target wants to requeue the I/O */ - dm_requeue_original_request(md, tio->orig); - break; - default: - if (r > 0) { - DMWARN("unimplemented target map return value: %d", r); - BUG(); - } - - /* The target wants to complete the I/O */ - dm_kill_unmapped_request(rq, r); - return r; - } - - return 0; -} - -static void map_tio_request(struct kthread_work *work) -{ - struct dm_rq_target_io *tio = container_of(work, struct dm_rq_target_io, work); - struct request *rq = tio->orig; - struct mapped_device *md = tio->md; - - if (map_request(tio, rq, md) == DM_MAPIO_REQUEUE) - dm_requeue_original_request(md, rq); -} - -static void dm_start_request(struct mapped_device *md, struct request *orig) -{ - if (!orig->q->mq_ops) - blk_start_request(orig); - else - blk_mq_start_request(orig); - atomic_inc(&md->pending[rq_data_dir(orig)]); - - if (md->seq_rq_merge_deadline_usecs) { - md->last_rq_pos = rq_end_sector(orig); - md->last_rq_rw = rq_data_dir(orig); - md->last_rq_start_time = ktime_get(); - } - - if (unlikely(dm_stats_used(&md->stats))) { - struct dm_rq_target_io *tio = tio_from_request(orig); - tio->duration_jiffies = jiffies; - tio->n_sectors = blk_rq_sectors(orig); - dm_stats_account_io(&md->stats, rq_data_dir(orig), - blk_rq_pos(orig), tio->n_sectors, false, 0, - &tio->stats_aux); - } - - /* - * Hold the md reference here for the in-flight I/O. - * We can't rely on the reference count by device opener, - * because the device may be closed during the request completion - * when all bios are completed. - * See the comment in rq_completed() too. - */ - dm_get(md); -} - -#define MAX_SEQ_RQ_MERGE_DEADLINE_USECS 100000 - -ssize_t dm_attr_rq_based_seq_io_merge_deadline_show(struct mapped_device *md, char *buf) -{ - return sprintf(buf, "%u\n", md->seq_rq_merge_deadline_usecs); -} - -ssize_t dm_attr_rq_based_seq_io_merge_deadline_store(struct mapped_device *md, - const char *buf, size_t count) -{ - unsigned deadline; - - if (!dm_request_based(md) || md->use_blk_mq) - return count; - - if (kstrtouint(buf, 10, &deadline)) - return -EINVAL; - - if (deadline > MAX_SEQ_RQ_MERGE_DEADLINE_USECS) - deadline = MAX_SEQ_RQ_MERGE_DEADLINE_USECS; - - md->seq_rq_merge_deadline_usecs = deadline; - - return count; -} - -static bool dm_request_peeked_before_merge_deadline(struct mapped_device *md) -{ - ktime_t kt_deadline; - - if (!md->seq_rq_merge_deadline_usecs) - return false; - - kt_deadline = ns_to_ktime((u64)md->seq_rq_merge_deadline_usecs * NSEC_PER_USEC); - kt_deadline = ktime_add_safe(md->last_rq_start_time, kt_deadline); - - return !ktime_after(ktime_get(), kt_deadline); -} - -/* - * q->request_fn for request-based dm. - * Called with the queue lock held. - */ -static void dm_request_fn(struct request_queue *q) -{ - struct mapped_device *md = q->queuedata; - struct dm_target *ti = md->immutable_target; - struct request *rq; - struct dm_rq_target_io *tio; - sector_t pos = 0; - - if (unlikely(!ti)) { - int srcu_idx; - struct dm_table *map = dm_get_live_table(md, &srcu_idx); - - ti = dm_table_find_target(map, pos); - dm_put_live_table(md, srcu_idx); - } - - /* - * For suspend, check blk_queue_stopped() and increment - * ->pending within a single queue_lock not to increment the - * number of in-flight I/Os after the queue is stopped in - * dm_suspend(). - */ - while (!blk_queue_stopped(q)) { - rq = blk_peek_request(q); - if (!rq) - return; - - /* always use block 0 to find the target for flushes for now */ - pos = 0; - if (req_op(rq) != REQ_OP_FLUSH) - pos = blk_rq_pos(rq); - - if ((dm_request_peeked_before_merge_deadline(md) && - md_in_flight(md) && rq->bio && rq->bio->bi_vcnt == 1 && - md->last_rq_pos == pos && md->last_rq_rw == rq_data_dir(rq)) || - (ti->type->busy && ti->type->busy(ti))) { - blk_delay_queue(q, HZ / 100); - return; - } - - dm_start_request(md, rq); - - tio = tio_from_request(rq); - /* Establish tio->ti before queuing work (map_tio_request) */ - tio->ti = ti; - queue_kthread_work(&md->kworker, &tio->work); - BUG_ON(!irqs_disabled()); - } -} - static int dm_any_congested(void *congested_data, int bdi_bits) { int r = bdi_bits; @@ -2269,7 +1379,7 @@ static const struct block_device_operations dm_blk_dops; static void dm_wq_work(struct work_struct *work); -static void dm_init_md_queue(struct mapped_device *md) +void dm_init_md_queue(struct mapped_device *md) { /* * Request-based dm devices cannot be stacked on top of bio-based dm @@ -2290,7 +1400,7 @@ static void dm_init_md_queue(struct mapped_device *md) md->queue->backing_dev_info.congested_data = md; } -static void dm_init_normal_md_queue(struct mapped_device *md) +void dm_init_normal_md_queue(struct mapped_device *md) { md->use_blk_mq = false; dm_init_md_queue(md); @@ -2330,6 +1440,8 @@ static void cleanup_mapped_device(struct mapped_device *md) bdput(md->bdev); md->bdev = NULL; } + + dm_mq_cleanup_mapped_device(md); } /* @@ -2363,7 +1475,7 @@ static struct mapped_device *alloc_dev(int minor) goto bad_io_barrier; md->numa_node_id = numa_node_id; - md->use_blk_mq = use_blk_mq; + md->use_blk_mq = dm_use_blk_mq_default(); md->init_tio_pdu = false; md->type = DM_TYPE_NONE; mutex_init(&md->suspend_lock); @@ -2448,10 +1560,6 @@ static void free_dev(struct mapped_device *md) unlock_fs(md); cleanup_mapped_device(md); - if (md->tag_set) { - blk_mq_free_tag_set(md->tag_set); - kfree(md->tag_set); - } free_table_devices(&md->table_devices); dm_stats_cleanup(&md->stats); @@ -2467,7 +1575,7 @@ static void __bind_mempools(struct mapped_device *md, struct dm_table *t) if (md->bs) { /* The md already has necessary mempools. */ - if (dm_table_get_type(t) == DM_TYPE_BIO_BASED) { + if (dm_table_bio_based(t)) { /* * Reload bioset because front_pad may have changed * because a different table was loaded. @@ -2657,176 +1765,15 @@ struct queue_limits *dm_get_queue_limits(struct mapped_device *md) } EXPORT_SYMBOL_GPL(dm_get_queue_limits); -static void dm_old_init_rq_based_worker_thread(struct mapped_device *md) -{ - /* Initialize the request-based DM worker thread */ - init_kthread_worker(&md->kworker); - md->kworker_task = kthread_run(kthread_worker_fn, &md->kworker, - "kdmwork-%s", dm_device_name(md)); -} - -/* - * Fully initialize a .request_fn request-based queue. - */ -static int dm_old_init_request_queue(struct mapped_device *md) -{ - /* Fully initialize the queue */ - if (!blk_init_allocated_queue(md->queue, dm_request_fn, NULL)) - return -EINVAL; - - /* disable dm_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); - blk_queue_prep_rq(md->queue, dm_old_prep_fn); - - dm_old_init_rq_based_worker_thread(md); - - elv_register_queue(md->queue); - - return 0; -} - -static int dm_mq_init_request(void *data, struct request *rq, - unsigned int hctx_idx, unsigned int request_idx, - unsigned int numa_node) -{ - struct mapped_device *md = data; - struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq); - - /* - * Must initialize md member of tio, otherwise it won't - * be available in dm_mq_queue_rq. - */ - tio->md = md; - - if (md->init_tio_pdu) { - /* target-specific per-io data is immediately after the tio */ - tio->info.ptr = tio + 1; - } - - return 0; -} - -static int dm_mq_queue_rq(struct blk_mq_hw_ctx *hctx, - const struct blk_mq_queue_data *bd) -{ - struct request *rq = bd->rq; - struct dm_rq_target_io *tio = blk_mq_rq_to_pdu(rq); - struct mapped_device *md = tio->md; - struct dm_target *ti = md->immutable_target; - - if (unlikely(!ti)) { - int srcu_idx; - struct dm_table *map = dm_get_live_table(md, &srcu_idx); - - ti = dm_table_find_target(map, 0); - dm_put_live_table(md, srcu_idx); - } - - if (ti->type->busy && ti->type->busy(ti)) - return BLK_MQ_RQ_QUEUE_BUSY; - - dm_start_request(md, rq); - - /* Init tio using md established in .init_request */ - init_tio(tio, rq, md); - - /* - * Establish tio->ti before queuing work (map_tio_request) - * or making direct call to map_request(). - */ - tio->ti = ti; - - /* Direct call is fine since .queue_rq allows allocations */ - if (map_request(tio, rq, md) == DM_MAPIO_REQUEUE) { - /* Undo dm_start_request() before requeuing */ - rq_end_stats(md, rq); - rq_completed(md, rq_data_dir(rq), false); - return BLK_MQ_RQ_QUEUE_BUSY; - } - - return BLK_MQ_RQ_QUEUE_OK; -} - -static struct blk_mq_ops dm_mq_ops = { - .queue_rq = dm_mq_queue_rq, - .map_queue = blk_mq_map_queue, - .complete = dm_softirq_done, - .init_request = dm_mq_init_request, -}; - -static int dm_mq_init_request_queue(struct mapped_device *md, - struct dm_target *immutable_tgt) -{ - struct request_queue *q; - int err; - - if (dm_get_md_type(md) == DM_TYPE_REQUEST_BASED) { - DMERR("request-based dm-mq may only be stacked on blk-mq device(s)"); - return -EINVAL; - } - - md->tag_set = kzalloc_node(sizeof(struct blk_mq_tag_set), GFP_KERNEL, md->numa_node_id); - if (!md->tag_set) - return -ENOMEM; - - md->tag_set->ops = &dm_mq_ops; - md->tag_set->queue_depth = dm_get_blk_mq_queue_depth(); - md->tag_set->numa_node = md->numa_node_id; - md->tag_set->flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_SG_MERGE; - md->tag_set->nr_hw_queues = dm_get_blk_mq_nr_hw_queues(); - md->tag_set->driver_data = md; - - md->tag_set->cmd_size = sizeof(struct dm_rq_target_io); - if (immutable_tgt && immutable_tgt->per_io_data_size) { - /* any target-specific per-io data is immediately after the tio */ - md->tag_set->cmd_size += immutable_tgt->per_io_data_size; - md->init_tio_pdu = true; - } - - err = blk_mq_alloc_tag_set(md->tag_set); - if (err) - goto out_kfree_tag_set; - - q = blk_mq_init_allocated_queue(md->tag_set, md->queue); - if (IS_ERR(q)) { - err = PTR_ERR(q); - goto out_tag_set; - } - dm_init_md_queue(md); - - /* backfill 'mq' sysfs registration normally done in blk_register_queue */ - blk_mq_register_disk(md->disk); - - return 0; - -out_tag_set: - blk_mq_free_tag_set(md->tag_set); -out_kfree_tag_set: - kfree(md->tag_set); - - return err; -} - -static unsigned filter_md_type(unsigned type, struct mapped_device *md) -{ - if (type == DM_TYPE_BIO_BASED) - return type; - - return !md->use_blk_mq ? DM_TYPE_REQUEST_BASED : DM_TYPE_MQ_REQUEST_BASED; -} - /* * Setup the DM device's queue based on md's type */ int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t) { int r; - unsigned md_type = filter_md_type(dm_get_md_type(md), md); + unsigned type = dm_get_md_type(md); - switch (md_type) { + switch (type) { case DM_TYPE_REQUEST_BASED: r = dm_old_init_request_queue(md); if (r) { @@ -2835,13 +1782,14 @@ int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t) } break; case DM_TYPE_MQ_REQUEST_BASED: - r = dm_mq_init_request_queue(md, dm_table_get_immutable_target(t)); + r = dm_mq_init_request_queue(md, t); if (r) { DMERR("Cannot initialize queue for request-based dm-mq mapped device"); return r; } break; case DM_TYPE_BIO_BASED: + case DM_TYPE_DAX_BIO_BASED: dm_init_normal_md_queue(md); blk_queue_make_request(md->queue, dm_make_request); /* @@ -2850,6 +1798,9 @@ int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t) */ 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; } @@ -3544,10 +2495,9 @@ struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, unsigned t if (!pools) return NULL; - type = filter_md_type(type, md); - switch (type) { case DM_TYPE_BIO_BASED: + case DM_TYPE_DAX_BIO_BASED: cachep = _io_cache; pool_size = dm_get_reserved_bio_based_ios(); front_pad = roundup(per_io_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone); @@ -3604,26 +2554,76 @@ void dm_free_md_mempools(struct dm_md_mempools *pools) kfree(pools); } -static int dm_pr_register(struct block_device *bdev, u64 old_key, u64 new_key, - u32 flags) +struct dm_pr { + u64 old_key; + u64 new_key; + u32 flags; + bool fail_early; +}; + +static int dm_call_pr(struct block_device *bdev, iterate_devices_callout_fn fn, + void *data) { struct mapped_device *md = bdev->bd_disk->private_data; - const struct pr_ops *ops; - fmode_t mode; - int r; + struct dm_table *table; + struct dm_target *ti; + int ret = -ENOTTY, srcu_idx; - r = dm_grab_bdev_for_ioctl(md, &bdev, &mode); - if (r < 0) - return r; + table = dm_get_live_table(md, &srcu_idx); + if (!table || !dm_table_get_size(table)) + goto out; - ops = bdev->bd_disk->fops->pr_ops; - if (ops && ops->pr_register) - r = ops->pr_register(bdev, old_key, new_key, flags); - else - r = -EOPNOTSUPP; + /* We only support devices that have a single target */ + if (dm_table_get_num_targets(table) != 1) + goto out; + ti = dm_table_get_target(table, 0); - bdput(bdev); - return r; + ret = -EINVAL; + if (!ti->type->iterate_devices) + goto out; + + ret = ti->type->iterate_devices(ti, fn, data); +out: + dm_put_live_table(md, srcu_idx); + return ret; +} + +/* + * For register / unregister we need to manually call out to every path. + */ +static int __dm_pr_register(struct dm_target *ti, struct dm_dev *dev, + sector_t start, sector_t len, void *data) +{ + struct dm_pr *pr = data; + const struct pr_ops *ops = dev->bdev->bd_disk->fops->pr_ops; + + if (!ops || !ops->pr_register) + return -EOPNOTSUPP; + return ops->pr_register(dev->bdev, pr->old_key, pr->new_key, pr->flags); +} + +static int dm_pr_register(struct block_device *bdev, u64 old_key, u64 new_key, + u32 flags) +{ + struct dm_pr pr = { + .old_key = old_key, + .new_key = new_key, + .flags = flags, + .fail_early = true, + }; + int ret; + + ret = dm_call_pr(bdev, __dm_pr_register, &pr); + if (ret && new_key) { + /* unregister all paths if we failed to register any path */ + pr.old_key = new_key; + pr.new_key = 0; + pr.flags = 0; + pr.fail_early = false; + dm_call_pr(bdev, __dm_pr_register, &pr); + } + + return ret; } static int dm_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type, @@ -3724,6 +2724,7 @@ static const struct block_device_operations dm_blk_dops = { .open = dm_blk_open, .release = dm_blk_close, .ioctl = dm_blk_ioctl, + .direct_access = dm_blk_direct_access, .getgeo = dm_blk_getgeo, .pr_ops = &dm_pr_ops, .owner = THIS_MODULE @@ -3741,18 +2742,6 @@ MODULE_PARM_DESC(major, "The major number of the device mapper"); module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR); MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools"); -module_param(reserved_rq_based_ios, uint, S_IRUGO | S_IWUSR); -MODULE_PARM_DESC(reserved_rq_based_ios, "Reserved IOs in request-based mempools"); - -module_param(use_blk_mq, bool, S_IRUGO | S_IWUSR); -MODULE_PARM_DESC(use_blk_mq, "Use block multiqueue for request-based DM devices"); - -module_param(dm_mq_nr_hw_queues, uint, S_IRUGO | S_IWUSR); -MODULE_PARM_DESC(dm_mq_nr_hw_queues, "Number of hardware queues for request-based dm-mq devices"); - -module_param(dm_mq_queue_depth, uint, S_IRUGO | S_IWUSR); -MODULE_PARM_DESC(dm_mq_queue_depth, "Queue depth for request-based dm-mq devices"); - module_param(dm_numa_node, int, S_IRUGO | S_IWUSR); MODULE_PARM_DESC(dm_numa_node, "NUMA node for DM device memory allocations"); diff --git a/drivers/md/dm.h b/drivers/md/dm.h index 13a758ec0f88..f0aad08b9654 100644 --- a/drivers/md/dm.h +++ b/drivers/md/dm.h @@ -13,6 +13,7 @@ #include <linux/fs.h> #include <linux/device-mapper.h> #include <linux/list.h> +#include <linux/moduleparam.h> #include <linux/blkdev.h> #include <linux/backing-dev.h> #include <linux/hdreg.h> @@ -33,14 +34,6 @@ #define DM_STATUS_NOFLUSH_FLAG (1 << 0) /* - * Type of table and mapped_device's mempool - */ -#define DM_TYPE_NONE 0 -#define DM_TYPE_BIO_BASED 1 -#define DM_TYPE_REQUEST_BASED 2 -#define DM_TYPE_MQ_REQUEST_BASED 3 - -/* * List of devices that a metadevice uses and should open/close. */ struct dm_dev_internal { @@ -75,8 +68,9 @@ unsigned dm_table_get_type(struct dm_table *t); struct target_type *dm_table_get_immutable_target_type(struct dm_table *t); struct dm_target *dm_table_get_immutable_target(struct dm_table *t); struct dm_target *dm_table_get_wildcard_target(struct dm_table *t); +bool dm_table_bio_based(struct dm_table *t); bool dm_table_request_based(struct dm_table *t); -bool dm_table_mq_request_based(struct dm_table *t); +bool dm_table_all_blk_mq_devices(struct dm_table *t); void dm_table_free_md_mempools(struct dm_table *t); struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t); @@ -161,16 +155,6 @@ void dm_interface_exit(void); /* * sysfs interface */ -struct dm_kobject_holder { - struct kobject kobj; - struct completion completion; -}; - -static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj) -{ - return &container_of(kobj, struct dm_kobject_holder, kobj)->completion; -} - int dm_sysfs_init(struct mapped_device *md); void dm_sysfs_exit(struct mapped_device *md); struct kobject *dm_kobject(struct mapped_device *md); @@ -212,8 +196,6 @@ int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action, void dm_internal_suspend(struct mapped_device *md); void dm_internal_resume(struct mapped_device *md); -bool dm_use_blk_mq(struct mapped_device *md); - int dm_io_init(void); void dm_io_exit(void); @@ -228,18 +210,8 @@ struct dm_md_mempools *dm_alloc_md_mempools(struct mapped_device *md, unsigned t void dm_free_md_mempools(struct dm_md_mempools *pools); /* - * Helpers that are used by DM core + * Various helpers */ unsigned dm_get_reserved_bio_based_ios(void); -unsigned dm_get_reserved_rq_based_ios(void); - -static inline bool dm_message_test_buffer_overflow(char *result, unsigned maxlen) -{ - return !maxlen || strlen(result) + 1 >= maxlen; -} - -ssize_t dm_attr_rq_based_seq_io_merge_deadline_show(struct mapped_device *md, char *buf); -ssize_t dm_attr_rq_based_seq_io_merge_deadline_store(struct mapped_device *md, - const char *buf, size_t count); #endif diff --git a/drivers/md/persistent-data/dm-btree.c b/drivers/md/persistent-data/dm-btree.c index ea3d3b656fd0..2cc1877804c2 100644 --- a/drivers/md/persistent-data/dm-btree.c +++ b/drivers/md/persistent-data/dm-btree.c @@ -429,7 +429,14 @@ static int dm_btree_lookup_next_single(struct dm_btree_info *info, dm_block_t ro if (flags & INTERNAL_NODE) { i = lower_bound(n, key); - if (i < 0 || i >= nr_entries) { + if (i < 0) { + /* + * avoid early -ENODATA return when all entries are + * higher than the search @key. + */ + i = 0; + } + if (i >= nr_entries) { r = -ENODATA; goto out; } diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c index fa85d19b81a6..d3e852ad5aa3 100644 --- a/drivers/scsi/sd.c +++ b/drivers/scsi/sd.c @@ -1619,8 +1619,7 @@ static int sd_pr_register(struct block_device *bdev, u64 old_key, u64 new_key, return -EOPNOTSUPP; return sd_pr_command(bdev, (flags & PR_FL_IGNORE_KEY) ? 0x06 : 0x00, old_key, new_key, 0, - (1 << 0) /* APTPL */ | - (1 << 2) /* ALL_TG_PT */); + (1 << 0) /* APTPL */); } static int sd_pr_reserve(struct block_device *bdev, u64 key, enum pr_type type, diff --git a/include/linux/device-mapper.h b/include/linux/device-mapper.h index 0830c9e86f0d..b0db857f334b 100644 --- a/include/linux/device-mapper.h +++ b/include/linux/device-mapper.h @@ -19,6 +19,15 @@ struct dm_table; struct mapped_device; struct bio_vec; +/* + * Type of table, mapped_device's mempool and request_queue + */ +#define DM_TYPE_NONE 0 +#define DM_TYPE_BIO_BASED 1 +#define DM_TYPE_REQUEST_BASED 2 +#define DM_TYPE_MQ_REQUEST_BASED 3 +#define DM_TYPE_DAX_BIO_BASED 4 + typedef enum { STATUSTYPE_INFO, STATUSTYPE_TABLE } status_type_t; union map_info { @@ -116,6 +125,14 @@ typedef void (*dm_io_hints_fn) (struct dm_target *ti, */ typedef int (*dm_busy_fn) (struct dm_target *ti); +/* + * Returns: + * < 0 : error + * >= 0 : the number of bytes accessible at the address + */ +typedef long (*dm_direct_access_fn) (struct dm_target *ti, sector_t sector, + void __pmem **kaddr, pfn_t *pfn, long size); + void dm_error(const char *message); struct dm_dev { @@ -162,6 +179,7 @@ struct target_type { dm_busy_fn busy; dm_iterate_devices_fn iterate_devices; dm_io_hints_fn io_hints; + dm_direct_access_fn direct_access; /* For internal device-mapper use. */ struct list_head list; @@ -444,6 +462,14 @@ int dm_table_add_target(struct dm_table *t, const char *type, void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb); /* + * Target can use this to set the table's type. + * Can only ever be called from a target's ctr. + * Useful for "hybrid" target (supports both bio-based + * and request-based). + */ +void dm_table_set_type(struct dm_table *t, unsigned type); + +/* * Finally call this to make the table ready for use. */ int dm_table_complete(struct dm_table *t); diff --git a/include/uapi/linux/dm-ioctl.h b/include/uapi/linux/dm-ioctl.h index 30afd0a23c4b..4bf9f1eabffc 100644 --- a/include/uapi/linux/dm-ioctl.h +++ b/include/uapi/linux/dm-ioctl.h @@ -267,9 +267,9 @@ enum { #define DM_DEV_SET_GEOMETRY _IOWR(DM_IOCTL, DM_DEV_SET_GEOMETRY_CMD, struct dm_ioctl) #define DM_VERSION_MAJOR 4 -#define DM_VERSION_MINOR 34 +#define DM_VERSION_MINOR 35 #define DM_VERSION_PATCHLEVEL 0 -#define DM_VERSION_EXTRA "-ioctl (2015-10-28)" +#define DM_VERSION_EXTRA "-ioctl (2016-06-23)" /* Status bits */ #define DM_READONLY_FLAG (1 << 0) /* In/Out */ |