diff options
Diffstat (limited to 'drivers/md')
32 files changed, 1153 insertions, 384 deletions
diff --git a/drivers/md/bcache/alloc.c b/drivers/md/bcache/alloc.c index a1df0d95151c..8bc1faf71ff2 100644 --- a/drivers/md/bcache/alloc.c +++ b/drivers/md/bcache/alloc.c @@ -67,6 +67,7 @@ #include <linux/blkdev.h> #include <linux/kthread.h> #include <linux/random.h> +#include <linux/sched/signal.h> #include <trace/events/bcache.h> #define MAX_OPEN_BUCKETS 128 @@ -733,8 +734,21 @@ int bch_open_buckets_alloc(struct cache_set *c) int bch_cache_allocator_start(struct cache *ca) { - struct task_struct *k = kthread_run(bch_allocator_thread, - ca, "bcache_allocator"); + struct task_struct *k; + + /* + * In case previous btree check operation occupies too many + * system memory for bcache btree node cache, and the + * registering process is selected by OOM killer. Here just + * ignore the SIGKILL sent by OOM killer if there is, to + * avoid kthread_run() being failed by pending signals. The + * bcache registering process will exit after the registration + * done. + */ + if (signal_pending(current)) + flush_signals(current); + + k = kthread_run(bch_allocator_thread, ca, "bcache_allocator"); if (IS_ERR(k)) return PTR_ERR(k); diff --git a/drivers/md/bcache/bcache.h b/drivers/md/bcache/bcache.h index 9198c1b480d9..74a9849ea164 100644 --- a/drivers/md/bcache/bcache.h +++ b/drivers/md/bcache/bcache.h @@ -301,6 +301,7 @@ struct cached_dev { struct block_device *bdev; struct cache_sb sb; + struct cache_sb_disk *sb_disk; struct bio sb_bio; struct bio_vec sb_bv[1]; struct closure sb_write; @@ -329,6 +330,9 @@ struct cached_dev { */ atomic_t has_dirty; +#define BCH_CACHE_READA_ALL 0 +#define BCH_CACHE_READA_META_ONLY 1 + unsigned int cache_readahead_policy; struct bch_ratelimit writeback_rate; struct delayed_work writeback_rate_update; @@ -403,6 +407,7 @@ enum alloc_reserve { struct cache { struct cache_set *set; struct cache_sb sb; + struct cache_sb_disk *sb_disk; struct bio sb_bio; struct bio_vec sb_bv[1]; diff --git a/drivers/md/bcache/bset.c b/drivers/md/bcache/bset.c index cffcdc9feefb..4385303836d8 100644 --- a/drivers/md/bcache/bset.c +++ b/drivers/md/bcache/bset.c @@ -1257,6 +1257,11 @@ static void __btree_sort(struct btree_keys *b, struct btree_iter *iter, * Our temporary buffer is the same size as the btree node's * buffer, we can just swap buffers instead of doing a big * memcpy() + * + * Don't worry event 'out' is allocated from mempool, it can + * still be swapped here. Because state->pool is a page mempool + * creaated by by mempool_init_page_pool(), which allocates + * pages by alloc_pages() indeed. */ out->magic = b->set->data->magic; diff --git a/drivers/md/bcache/bset.h b/drivers/md/bcache/bset.h index c71365e7c1fa..a50dcfda656f 100644 --- a/drivers/md/bcache/bset.h +++ b/drivers/md/bcache/bset.h @@ -397,7 +397,8 @@ void bch_btree_keys_stats(struct btree_keys *b, struct bset_stats *state); /* Bkey utility code */ -#define bset_bkey_last(i) bkey_idx((struct bkey *) (i)->d, (i)->keys) +#define bset_bkey_last(i) bkey_idx((struct bkey *) (i)->d, \ + (unsigned int)(i)->keys) static inline struct bkey *bset_bkey_idx(struct bset *i, unsigned int idx) { diff --git a/drivers/md/bcache/btree.c b/drivers/md/bcache/btree.c index 14d6c33b0957..b12186c87f52 100644 --- a/drivers/md/bcache/btree.c +++ b/drivers/md/bcache/btree.c @@ -34,6 +34,7 @@ #include <linux/random.h> #include <linux/rcupdate.h> #include <linux/sched/clock.h> +#include <linux/sched/signal.h> #include <linux/rculist.h> #include <linux/delay.h> #include <trace/events/bcache.h> @@ -734,34 +735,32 @@ static unsigned long bch_mca_scan(struct shrinker *shrink, i = 0; btree_cache_used = c->btree_cache_used; - list_for_each_entry_safe(b, t, &c->btree_cache_freeable, list) { + list_for_each_entry_safe_reverse(b, t, &c->btree_cache_freeable, list) { if (nr <= 0) goto out; - if (++i > 3 && - !mca_reap(b, 0, false)) { + if (!mca_reap(b, 0, false)) { mca_data_free(b); rw_unlock(true, b); freed++; } nr--; + i++; } - for (; (nr--) && i < btree_cache_used; i++) { - if (list_empty(&c->btree_cache)) + list_for_each_entry_safe_reverse(b, t, &c->btree_cache, list) { + if (nr <= 0 || i >= btree_cache_used) goto out; - b = list_first_entry(&c->btree_cache, struct btree, list); - list_rotate_left(&c->btree_cache); - - if (!b->accessed && - !mca_reap(b, 0, false)) { + if (!mca_reap(b, 0, false)) { mca_bucket_free(b); mca_data_free(b); rw_unlock(true, b); freed++; - } else - b->accessed = 0; + } + + nr--; + i++; } out: mutex_unlock(&c->bucket_lock); @@ -1069,7 +1068,6 @@ retry: BUG_ON(!b->written); b->parent = parent; - b->accessed = 1; for (; i <= b->keys.nsets && b->keys.set[i].size; i++) { prefetch(b->keys.set[i].tree); @@ -1160,7 +1158,6 @@ retry: goto retry; } - b->accessed = 1; b->parent = parent; bch_bset_init_next(&b->keys, b->keys.set->data, bset_magic(&b->c->sb)); @@ -1917,6 +1914,18 @@ static int bch_gc_thread(void *arg) int bch_gc_thread_start(struct cache_set *c) { + /* + * In case previous btree check operation occupies too many + * system memory for bcache btree node cache, and the + * registering process is selected by OOM killer. Here just + * ignore the SIGKILL sent by OOM killer if there is, to + * avoid kthread_run() being failed by pending signals. The + * bcache registering process will exit after the registration + * done. + */ + if (signal_pending(current)) + flush_signals(current); + c->gc_thread = kthread_run(bch_gc_thread, c, "bcache_gc"); return PTR_ERR_OR_ZERO(c->gc_thread); } diff --git a/drivers/md/bcache/btree.h b/drivers/md/bcache/btree.h index 76cfd121a486..f4dcca449391 100644 --- a/drivers/md/bcache/btree.h +++ b/drivers/md/bcache/btree.h @@ -121,8 +121,6 @@ struct btree { /* Key/pointer for this btree node */ BKEY_PADDED(key); - /* Single bit - set when accessed, cleared by shrinker */ - unsigned long accessed; unsigned long seq; struct rw_semaphore lock; struct cache_set *c; diff --git a/drivers/md/bcache/journal.c b/drivers/md/bcache/journal.c index be2a2a201603..0e3ff9745ac7 100644 --- a/drivers/md/bcache/journal.c +++ b/drivers/md/bcache/journal.c @@ -420,7 +420,10 @@ err: static void btree_flush_write(struct cache_set *c) { struct btree *b, *t, *btree_nodes[BTREE_FLUSH_NR]; - unsigned int i, n; + unsigned int i, nr; + int ref_nr; + atomic_t *fifo_front_p, *now_fifo_front_p; + size_t mask; if (c->journal.btree_flushing) return; @@ -433,12 +436,50 @@ static void btree_flush_write(struct cache_set *c) c->journal.btree_flushing = true; spin_unlock(&c->journal.flush_write_lock); + /* get the oldest journal entry and check its refcount */ + spin_lock(&c->journal.lock); + fifo_front_p = &fifo_front(&c->journal.pin); + ref_nr = atomic_read(fifo_front_p); + if (ref_nr <= 0) { + /* + * do nothing if no btree node references + * the oldest journal entry + */ + spin_unlock(&c->journal.lock); + goto out; + } + spin_unlock(&c->journal.lock); + + mask = c->journal.pin.mask; + nr = 0; atomic_long_inc(&c->flush_write); memset(btree_nodes, 0, sizeof(btree_nodes)); - n = 0; mutex_lock(&c->bucket_lock); list_for_each_entry_safe_reverse(b, t, &c->btree_cache, list) { + /* + * It is safe to get now_fifo_front_p without holding + * c->journal.lock here, because we don't need to know + * the exactly accurate value, just check whether the + * front pointer of c->journal.pin is changed. + */ + now_fifo_front_p = &fifo_front(&c->journal.pin); + /* + * If the oldest journal entry is reclaimed and front + * pointer of c->journal.pin changes, it is unnecessary + * to scan c->btree_cache anymore, just quit the loop and + * flush out what we have already. + */ + if (now_fifo_front_p != fifo_front_p) + break; + /* + * quit this loop if all matching btree nodes are + * scanned and record in btree_nodes[] already. + */ + ref_nr = atomic_read(fifo_front_p); + if (nr >= ref_nr) + break; + if (btree_node_journal_flush(b)) pr_err("BUG: flush_write bit should not be set here!"); @@ -454,17 +495,43 @@ static void btree_flush_write(struct cache_set *c) continue; } + /* + * Only select the btree node which exactly references + * the oldest journal entry. + * + * If the journal entry pointed by fifo_front_p is + * reclaimed in parallel, don't worry: + * - the list_for_each_xxx loop will quit when checking + * next now_fifo_front_p. + * - If there are matched nodes recorded in btree_nodes[], + * they are clean now (this is why and how the oldest + * journal entry can be reclaimed). These selected nodes + * will be ignored and skipped in the folowing for-loop. + */ + if (((btree_current_write(b)->journal - fifo_front_p) & + mask) != 0) { + mutex_unlock(&b->write_lock); + continue; + } + set_btree_node_journal_flush(b); mutex_unlock(&b->write_lock); - btree_nodes[n++] = b; - if (n == BTREE_FLUSH_NR) + btree_nodes[nr++] = b; + /* + * To avoid holding c->bucket_lock too long time, + * only scan for BTREE_FLUSH_NR matched btree nodes + * at most. If there are more btree nodes reference + * the oldest journal entry, try to flush them next + * time when btree_flush_write() is called. + */ + if (nr == BTREE_FLUSH_NR) break; } mutex_unlock(&c->bucket_lock); - for (i = 0; i < n; i++) { + for (i = 0; i < nr; i++) { b = btree_nodes[i]; if (!b) { pr_err("BUG: btree_nodes[%d] is NULL", i); @@ -497,6 +564,7 @@ static void btree_flush_write(struct cache_set *c) mutex_unlock(&b->write_lock); } +out: spin_lock(&c->journal.flush_write_lock); c->journal.btree_flushing = false; spin_unlock(&c->journal.flush_write_lock); diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c index 73478a91a342..820d8402a1dc 100644 --- a/drivers/md/bcache/request.c +++ b/drivers/md/bcache/request.c @@ -379,13 +379,20 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio) goto skip; /* - * Flag for bypass if the IO is for read-ahead or background, - * unless the read-ahead request is for metadata + * If the bio is for read-ahead or background IO, bypass it or + * not depends on the following situations, + * - If the IO is for meta data, always cache it and no bypass + * - If the IO is not meta data, check dc->cache_reada_policy, + * BCH_CACHE_READA_ALL: cache it and not bypass + * BCH_CACHE_READA_META_ONLY: not cache it and bypass + * That is, read-ahead request for metadata always get cached * (eg, for gfs2 or xfs). */ - if (bio->bi_opf & (REQ_RAHEAD|REQ_BACKGROUND) && - !(bio->bi_opf & (REQ_META|REQ_PRIO))) - goto skip; + if ((bio->bi_opf & (REQ_RAHEAD|REQ_BACKGROUND))) { + if (!(bio->bi_opf & (REQ_META|REQ_PRIO)) && + (dc->cache_readahead_policy != BCH_CACHE_READA_ALL)) + goto skip; + } if (bio->bi_iter.bi_sector & (c->sb.block_size - 1) || bio_sectors(bio) & (c->sb.block_size - 1)) { diff --git a/drivers/md/bcache/stats.c b/drivers/md/bcache/stats.c index ba1c93791d8d..503aafe188dc 100644 --- a/drivers/md/bcache/stats.c +++ b/drivers/md/bcache/stats.c @@ -109,9 +109,13 @@ int bch_cache_accounting_add_kobjs(struct cache_accounting *acc, void bch_cache_accounting_clear(struct cache_accounting *acc) { - memset(&acc->total.cache_hits, - 0, - sizeof(struct cache_stats)); + acc->total.cache_hits = 0; + acc->total.cache_misses = 0; + acc->total.cache_bypass_hits = 0; + acc->total.cache_bypass_misses = 0; + acc->total.cache_readaheads = 0; + acc->total.cache_miss_collisions = 0; + acc->total.sectors_bypassed = 0; } void bch_cache_accounting_destroy(struct cache_accounting *acc) diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c index 77e9869345e7..0c3c5419c52b 100644 --- a/drivers/md/bcache/super.c +++ b/drivers/md/bcache/super.c @@ -15,7 +15,6 @@ #include "writeback.h" #include <linux/blkdev.h> -#include <linux/buffer_head.h> #include <linux/debugfs.h> #include <linux/genhd.h> #include <linux/idr.h> @@ -60,17 +59,18 @@ struct workqueue_struct *bch_journal_wq; /* Superblock */ static const char *read_super(struct cache_sb *sb, struct block_device *bdev, - struct page **res) + struct cache_sb_disk **res) { const char *err; - struct cache_sb *s; - struct buffer_head *bh = __bread(bdev, 1, SB_SIZE); + struct cache_sb_disk *s; + struct page *page; unsigned int i; - if (!bh) + page = read_cache_page_gfp(bdev->bd_inode->i_mapping, + SB_OFFSET >> PAGE_SHIFT, GFP_KERNEL); + if (IS_ERR(page)) return "IO error"; - - s = (struct cache_sb *) bh->b_data; + s = page_address(page) + offset_in_page(SB_OFFSET); sb->offset = le64_to_cpu(s->offset); sb->version = le64_to_cpu(s->version); @@ -188,12 +188,10 @@ static const char *read_super(struct cache_sb *sb, struct block_device *bdev, } sb->last_mount = (u32)ktime_get_real_seconds(); - err = NULL; - - get_page(bh->b_page); - *res = bh->b_page; + *res = s; + return NULL; err: - put_bh(bh); + put_page(page); return err; } @@ -207,15 +205,15 @@ static void write_bdev_super_endio(struct bio *bio) closure_put(&dc->sb_write); } -static void __write_super(struct cache_sb *sb, struct bio *bio) +static void __write_super(struct cache_sb *sb, struct cache_sb_disk *out, + struct bio *bio) { - struct cache_sb *out = page_address(bio_first_page_all(bio)); unsigned int i; + bio->bi_opf = REQ_OP_WRITE | REQ_SYNC | REQ_META; bio->bi_iter.bi_sector = SB_SECTOR; - bio->bi_iter.bi_size = SB_SIZE; - bio_set_op_attrs(bio, REQ_OP_WRITE, REQ_SYNC|REQ_META); - bch_bio_map(bio, NULL); + __bio_add_page(bio, virt_to_page(out), SB_SIZE, + offset_in_page(out)); out->offset = cpu_to_le64(sb->offset); out->version = cpu_to_le64(sb->version); @@ -257,14 +255,14 @@ void bch_write_bdev_super(struct cached_dev *dc, struct closure *parent) down(&dc->sb_write_mutex); closure_init(cl, parent); - bio_reset(bio); + bio_init(bio, dc->sb_bv, 1); bio_set_dev(bio, dc->bdev); bio->bi_end_io = write_bdev_super_endio; bio->bi_private = dc; closure_get(cl); /* I/O request sent to backing device */ - __write_super(&dc->sb, bio); + __write_super(&dc->sb, dc->sb_disk, bio); closure_return_with_destructor(cl, bch_write_bdev_super_unlock); } @@ -306,13 +304,13 @@ void bcache_write_super(struct cache_set *c) SET_CACHE_SYNC(&ca->sb, CACHE_SYNC(&c->sb)); - bio_reset(bio); + bio_init(bio, ca->sb_bv, 1); bio_set_dev(bio, ca->bdev); bio->bi_end_io = write_super_endio; bio->bi_private = ca; closure_get(cl); - __write_super(&ca->sb, bio); + __write_super(&ca->sb, ca->sb_disk, bio); } closure_return_with_destructor(cl, bcache_write_super_unlock); @@ -611,12 +609,13 @@ int bch_prio_write(struct cache *ca, bool wait) return 0; } -static void prio_read(struct cache *ca, uint64_t bucket) +static int prio_read(struct cache *ca, uint64_t bucket) { struct prio_set *p = ca->disk_buckets; struct bucket_disk *d = p->data + prios_per_bucket(ca), *end = d; struct bucket *b; unsigned int bucket_nr = 0; + int ret = -EIO; for (b = ca->buckets; b < ca->buckets + ca->sb.nbuckets; @@ -629,11 +628,15 @@ static void prio_read(struct cache *ca, uint64_t bucket) prio_io(ca, bucket, REQ_OP_READ, 0); if (p->csum != - bch_crc64(&p->magic, bucket_bytes(ca) - 8)) + bch_crc64(&p->magic, bucket_bytes(ca) - 8)) { pr_warn("bad csum reading priorities"); + goto out; + } - if (p->magic != pset_magic(&ca->sb)) + if (p->magic != pset_magic(&ca->sb)) { pr_warn("bad magic reading priorities"); + goto out; + } bucket = p->next_bucket; d = p->data; @@ -642,6 +645,10 @@ static void prio_read(struct cache *ca, uint64_t bucket) b->prio = le16_to_cpu(d->prio); b->gen = b->last_gc = d->gen; } + + ret = 0; +out: + return ret; } /* Bcache device */ @@ -1275,6 +1282,9 @@ static void cached_dev_free(struct closure *cl) mutex_unlock(&bch_register_lock); + if (dc->sb_disk) + put_page(virt_to_page(dc->sb_disk)); + if (!IS_ERR_OR_NULL(dc->bdev)) blkdev_put(dc->bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); @@ -1350,7 +1360,7 @@ static int cached_dev_init(struct cached_dev *dc, unsigned int block_size) /* Cached device - bcache superblock */ -static int register_bdev(struct cache_sb *sb, struct page *sb_page, +static int register_bdev(struct cache_sb *sb, struct cache_sb_disk *sb_disk, struct block_device *bdev, struct cached_dev *dc) { @@ -1362,11 +1372,7 @@ static int register_bdev(struct cache_sb *sb, struct page *sb_page, memcpy(&dc->sb, sb, sizeof(struct cache_sb)); dc->bdev = bdev; dc->bdev->bd_holder = dc; - - bio_init(&dc->sb_bio, dc->sb_bio.bi_inline_vecs, 1); - bio_first_bvec_all(&dc->sb_bio)->bv_page = sb_page; - get_page(sb_page); - + dc->sb_disk = sb_disk; if (cached_dev_init(dc, sb->block_size << 9)) goto err; @@ -1876,8 +1882,10 @@ static int run_cache_set(struct cache_set *c) j = &list_entry(journal.prev, struct journal_replay, list)->j; err = "IO error reading priorities"; - for_each_cache(ca, c, i) - prio_read(ca, j->prio_bucket[ca->sb.nr_this_dev]); + for_each_cache(ca, c, i) { + if (prio_read(ca, j->prio_bucket[ca->sb.nr_this_dev])) + goto err; + } /* * If prio_read() fails it'll call cache_set_error and we'll @@ -1909,23 +1917,6 @@ static int run_cache_set(struct cache_set *c) if (bch_btree_check(c)) goto err; - /* - * bch_btree_check() may occupy too much system memory which - * has negative effects to user space application (e.g. data - * base) performance. Shrink the mca cache memory proactively - * here to avoid competing memory with user space workloads.. - */ - if (!c->shrinker_disabled) { - struct shrink_control sc; - - sc.gfp_mask = GFP_KERNEL; - sc.nr_to_scan = c->btree_cache_used * c->btree_pages; - /* first run to clear b->accessed tag */ - c->shrink.scan_objects(&c->shrink, &sc); - /* second run to reap non-accessed nodes */ - c->shrink.scan_objects(&c->shrink, &sc); - } - bch_journal_mark(c, &journal); bch_initial_gc_finish(c); pr_debug("btree_check() done"); @@ -2136,8 +2127,8 @@ void bch_cache_release(struct kobject *kobj) for (i = 0; i < RESERVE_NR; i++) free_fifo(&ca->free[i]); - if (ca->sb_bio.bi_inline_vecs[0].bv_page) - put_page(bio_first_page_all(&ca->sb_bio)); + if (ca->sb_disk) + put_page(virt_to_page(ca->sb_disk)); if (!IS_ERR_OR_NULL(ca->bdev)) blkdev_put(ca->bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); @@ -2259,7 +2250,7 @@ err_free: return ret; } -static int register_cache(struct cache_sb *sb, struct page *sb_page, +static int register_cache(struct cache_sb *sb, struct cache_sb_disk *sb_disk, struct block_device *bdev, struct cache *ca) { const char *err = NULL; /* must be set for any error case */ @@ -2269,10 +2260,7 @@ static int register_cache(struct cache_sb *sb, struct page *sb_page, memcpy(&ca->sb, sb, sizeof(struct cache_sb)); ca->bdev = bdev; ca->bdev->bd_holder = ca; - - bio_init(&ca->sb_bio, ca->sb_bio.bi_inline_vecs, 1); - bio_first_bvec_all(&ca->sb_bio)->bv_page = sb_page; - get_page(sb_page); + ca->sb_disk = sb_disk; if (blk_queue_discard(bdev_get_queue(bdev))) ca->discard = CACHE_DISCARD(&ca->sb); @@ -2372,29 +2360,35 @@ static bool bch_is_open(struct block_device *bdev) static ssize_t register_bcache(struct kobject *k, struct kobj_attribute *attr, const char *buffer, size_t size) { - ssize_t ret = -EINVAL; - const char *err = "cannot allocate memory"; + const char *err; char *path = NULL; - struct cache_sb *sb = NULL; - struct block_device *bdev = NULL; - struct page *sb_page = NULL; + struct cache_sb *sb; + struct cache_sb_disk *sb_disk; + struct block_device *bdev; + ssize_t ret; + ret = -EBUSY; + err = "failed to reference bcache module"; if (!try_module_get(THIS_MODULE)) - return -EBUSY; + goto out; /* For latest state of bcache_is_reboot */ smp_mb(); + err = "bcache is in reboot"; if (bcache_is_reboot) - return -EBUSY; + goto out_module_put; + ret = -ENOMEM; + err = "cannot allocate memory"; path = kstrndup(buffer, size, GFP_KERNEL); if (!path) - goto err; + goto out_module_put; sb = kmalloc(sizeof(struct cache_sb), GFP_KERNEL); if (!sb) - goto err; + goto out_free_path; + ret = -EINVAL; err = "failed to open device"; bdev = blkdev_get_by_path(strim(path), FMODE_READ|FMODE_WRITE|FMODE_EXCL, @@ -2411,57 +2405,63 @@ static ssize_t register_bcache(struct kobject *k, struct kobj_attribute *attr, if (!IS_ERR(bdev)) bdput(bdev); if (attr == &ksysfs_register_quiet) - goto quiet_out; + goto done; } - goto err; + goto out_free_sb; } err = "failed to set blocksize"; if (set_blocksize(bdev, 4096)) - goto err_close; + goto out_blkdev_put; - err = read_super(sb, bdev, &sb_page); + err = read_super(sb, bdev, &sb_disk); if (err) - goto err_close; + goto out_blkdev_put; err = "failed to register device"; if (SB_IS_BDEV(sb)) { struct cached_dev *dc = kzalloc(sizeof(*dc), GFP_KERNEL); if (!dc) - goto err_close; + goto out_put_sb_page; mutex_lock(&bch_register_lock); - ret = register_bdev(sb, sb_page, bdev, dc); + ret = register_bdev(sb, sb_disk, bdev, dc); mutex_unlock(&bch_register_lock); /* blkdev_put() will be called in cached_dev_free() */ if (ret < 0) - goto err; + goto out_free_sb; } else { struct cache *ca = kzalloc(sizeof(*ca), GFP_KERNEL); if (!ca) - goto err_close; + goto out_put_sb_page; /* blkdev_put() will be called in bch_cache_release() */ - if (register_cache(sb, sb_page, bdev, ca) != 0) - goto err; + if (register_cache(sb, sb_disk, bdev, ca) != 0) + goto out_free_sb; } -quiet_out: - ret = size; -out: - if (sb_page) - put_page(sb_page); + +done: kfree(sb); kfree(path); module_put(THIS_MODULE); - return ret; + return size; -err_close: - blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); -err: - pr_info("error %s: %s", path, err); - goto out; +out_put_sb_page: + put_page(virt_to_page(sb_disk)); +out_blkdev_put: + blkdev_put(bdev, FMODE_READ | FMODE_WRITE | FMODE_EXCL); +out_free_sb: + kfree(sb); +out_free_path: + kfree(path); + path = NULL; +out_module_put: + module_put(THIS_MODULE); +out: + pr_info("error %s: %s", path?path:"", err); + return ret; } diff --git a/drivers/md/bcache/sysfs.c b/drivers/md/bcache/sysfs.c index 733e2ddf3c78..3470fae4eabc 100644 --- a/drivers/md/bcache/sysfs.c +++ b/drivers/md/bcache/sysfs.c @@ -27,6 +27,12 @@ static const char * const bch_cache_modes[] = { NULL }; +static const char * const bch_reada_cache_policies[] = { + "all", + "meta-only", + NULL +}; + /* Default is 0 ("auto") */ static const char * const bch_stop_on_failure_modes[] = { "auto", @@ -100,6 +106,7 @@ rw_attribute(congested_write_threshold_us); rw_attribute(sequential_cutoff); rw_attribute(data_csum); rw_attribute(cache_mode); +rw_attribute(readahead_cache_policy); rw_attribute(stop_when_cache_set_failed); rw_attribute(writeback_metadata); rw_attribute(writeback_running); @@ -168,6 +175,11 @@ SHOW(__bch_cached_dev) bch_cache_modes, BDEV_CACHE_MODE(&dc->sb)); + if (attr == &sysfs_readahead_cache_policy) + return bch_snprint_string_list(buf, PAGE_SIZE, + bch_reada_cache_policies, + dc->cache_readahead_policy); + if (attr == &sysfs_stop_when_cache_set_failed) return bch_snprint_string_list(buf, PAGE_SIZE, bch_stop_on_failure_modes, @@ -353,6 +365,15 @@ STORE(__cached_dev) } } + if (attr == &sysfs_readahead_cache_policy) { + v = __sysfs_match_string(bch_reada_cache_policies, -1, buf); + if (v < 0) + return v; + + if ((unsigned int) v != dc->cache_readahead_policy) + dc->cache_readahead_policy = v; + } + if (attr == &sysfs_stop_when_cache_set_failed) { v = __sysfs_match_string(bch_stop_on_failure_modes, -1, buf); if (v < 0) @@ -467,6 +488,7 @@ static struct attribute *bch_cached_dev_files[] = { &sysfs_data_csum, #endif &sysfs_cache_mode, + &sysfs_readahead_cache_policy, &sysfs_stop_when_cache_set_failed, &sysfs_writeback_metadata, &sysfs_writeback_running, diff --git a/drivers/md/dm-bio-prison-v2.c b/drivers/md/dm-bio-prison-v2.c index 8ee019eda32d..9dec3b61cf70 100644 --- a/drivers/md/dm-bio-prison-v2.c +++ b/drivers/md/dm-bio-prison-v2.c @@ -324,7 +324,7 @@ static bool __unlock(struct dm_bio_prison_v2 *prison, bio_list_init(&cell->bios); if (cell->shared_count) { - cell->exclusive_lock = 0; + cell->exclusive_lock = false; return false; } diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c index eb9782fc93fe..c6a529873d0f 100644 --- a/drivers/md/dm-crypt.c +++ b/drivers/md/dm-crypt.c @@ -1,8 +1,8 @@ /* * Copyright (C) 2003 Jana Saout <jana@saout.de> * Copyright (C) 2004 Clemens Fruhwirth <clemens@endorphin.org> - * Copyright (C) 2006-2017 Red Hat, Inc. All rights reserved. - * Copyright (C) 2013-2017 Milan Broz <gmazyland@gmail.com> + * Copyright (C) 2006-2020 Red Hat, Inc. All rights reserved. + * Copyright (C) 2013-2020 Milan Broz <gmazyland@gmail.com> * * This file is released under the GPL. */ @@ -115,6 +115,11 @@ struct iv_tcw_private { u8 *whitening; }; +#define ELEPHANT_MAX_KEY_SIZE 32 +struct iv_elephant_private { + struct crypto_skcipher *tfm; +}; + /* * Crypt: maps a linear range of a block device * and encrypts / decrypts at the same time. @@ -125,6 +130,7 @@ enum flags { DM_CRYPT_SUSPENDED, DM_CRYPT_KEY_VALID, enum cipher_flags { CRYPT_MODE_INTEGRITY_AEAD, /* Use authenticated mode for cihper */ CRYPT_IV_LARGE_SECTORS, /* Calculate IV from sector_size, not 512B sectors */ + CRYPT_ENCRYPT_PREPROCESS, /* Must preprocess data for encryption (elephant) */ }; /* @@ -152,6 +158,7 @@ struct crypt_config { struct iv_benbi_private benbi; struct iv_lmk_private lmk; struct iv_tcw_private tcw; + struct iv_elephant_private elephant; } iv_gen_private; u64 iv_offset; unsigned int iv_size; @@ -285,6 +292,11 @@ static struct crypto_aead *any_tfm_aead(struct crypt_config *cc) * eboiv: Encrypted byte-offset IV (used in Bitlocker in CBC mode) * The IV is encrypted little-endian byte-offset (with the same key * and cipher as the volume). + * + * elephant: The extended version of eboiv with additional Elephant diffuser + * used with Bitlocker CBC mode. + * This mode was used in older Windows systems + * http://download.microsoft.com/download/0/2/3/0238acaf-d3bf-4a6d-b3d6-0a0be4bbb36e/bitlockercipher200608.pdf */ static int crypt_iv_plain_gen(struct crypt_config *cc, u8 *iv, @@ -331,8 +343,14 @@ static int crypt_iv_essiv_gen(struct crypt_config *cc, u8 *iv, static int crypt_iv_benbi_ctr(struct crypt_config *cc, struct dm_target *ti, const char *opts) { - unsigned bs = crypto_skcipher_blocksize(any_tfm(cc)); - int log = ilog2(bs); + unsigned bs; + int log; + + if (test_bit(CRYPT_MODE_INTEGRITY_AEAD, &cc->cipher_flags)) + bs = crypto_aead_blocksize(any_tfm_aead(cc)); + else + bs = crypto_skcipher_blocksize(any_tfm(cc)); + log = ilog2(bs); /* we need to calculate how far we must shift the sector count * to get the cipher block count, we use this shift in _gen */ @@ -717,7 +735,7 @@ static int crypt_iv_eboiv_gen(struct crypt_config *cc, u8 *iv, struct crypto_wait wait; int err; - req = skcipher_request_alloc(any_tfm(cc), GFP_KERNEL | GFP_NOFS); + req = skcipher_request_alloc(any_tfm(cc), GFP_NOIO); if (!req) return -ENOMEM; @@ -734,6 +752,290 @@ static int crypt_iv_eboiv_gen(struct crypt_config *cc, u8 *iv, return err; } +static void crypt_iv_elephant_dtr(struct crypt_config *cc) +{ + struct iv_elephant_private *elephant = &cc->iv_gen_private.elephant; + + crypto_free_skcipher(elephant->tfm); + elephant->tfm = NULL; +} + +static int crypt_iv_elephant_ctr(struct crypt_config *cc, struct dm_target *ti, + const char *opts) +{ + struct iv_elephant_private *elephant = &cc->iv_gen_private.elephant; + int r; + + elephant->tfm = crypto_alloc_skcipher("ecb(aes)", 0, 0); + if (IS_ERR(elephant->tfm)) { + r = PTR_ERR(elephant->tfm); + elephant->tfm = NULL; + return r; + } + + r = crypt_iv_eboiv_ctr(cc, ti, NULL); + if (r) + crypt_iv_elephant_dtr(cc); + return r; +} + +static void diffuser_disk_to_cpu(u32 *d, size_t n) +{ +#ifndef __LITTLE_ENDIAN + int i; + + for (i = 0; i < n; i++) + d[i] = le32_to_cpu((__le32)d[i]); +#endif +} + +static void diffuser_cpu_to_disk(__le32 *d, size_t n) +{ +#ifndef __LITTLE_ENDIAN + int i; + + for (i = 0; i < n; i++) + d[i] = cpu_to_le32((u32)d[i]); +#endif +} + +static void diffuser_a_decrypt(u32 *d, size_t n) +{ + int i, i1, i2, i3; + + for (i = 0; i < 5; i++) { + i1 = 0; + i2 = n - 2; + i3 = n - 5; + + while (i1 < (n - 1)) { + d[i1] += d[i2] ^ (d[i3] << 9 | d[i3] >> 23); + i1++; i2++; i3++; + + if (i3 >= n) + i3 -= n; + + d[i1] += d[i2] ^ d[i3]; + i1++; i2++; i3++; + + if (i2 >= n) + i2 -= n; + + d[i1] += d[i2] ^ (d[i3] << 13 | d[i3] >> 19); + i1++; i2++; i3++; + + d[i1] += d[i2] ^ d[i3]; + i1++; i2++; i3++; + } + } +} + +static void diffuser_a_encrypt(u32 *d, size_t n) +{ + int i, i1, i2, i3; + + for (i = 0; i < 5; i++) { + i1 = n - 1; + i2 = n - 2 - 1; + i3 = n - 5 - 1; + + while (i1 > 0) { + d[i1] -= d[i2] ^ d[i3]; + i1--; i2--; i3--; + + d[i1] -= d[i2] ^ (d[i3] << 13 | d[i3] >> 19); + i1--; i2--; i3--; + + if (i2 < 0) + i2 += n; + + d[i1] -= d[i2] ^ d[i3]; + i1--; i2--; i3--; + + if (i3 < 0) + i3 += n; + + d[i1] -= d[i2] ^ (d[i3] << 9 | d[i3] >> 23); + i1--; i2--; i3--; + } + } +} + +static void diffuser_b_decrypt(u32 *d, size_t n) +{ + int i, i1, i2, i3; + + for (i = 0; i < 3; i++) { + i1 = 0; + i2 = 2; + i3 = 5; + + while (i1 < (n - 1)) { + d[i1] += d[i2] ^ d[i3]; + i1++; i2++; i3++; + + d[i1] += d[i2] ^ (d[i3] << 10 | d[i3] >> 22); + i1++; i2++; i3++; + + if (i2 >= n) + i2 -= n; + + d[i1] += d[i2] ^ d[i3]; + i1++; i2++; i3++; + + if (i3 >= n) + i3 -= n; + + d[i1] += d[i2] ^ (d[i3] << 25 | d[i3] >> 7); + i1++; i2++; i3++; + } + } +} + +static void diffuser_b_encrypt(u32 *d, size_t n) +{ + int i, i1, i2, i3; + + for (i = 0; i < 3; i++) { + i1 = n - 1; + i2 = 2 - 1; + i3 = 5 - 1; + + while (i1 > 0) { + d[i1] -= d[i2] ^ (d[i3] << 25 | d[i3] >> 7); + i1--; i2--; i3--; + + if (i3 < 0) + i3 += n; + + d[i1] -= d[i2] ^ d[i3]; + i1--; i2--; i3--; + + if (i2 < 0) + i2 += n; + + d[i1] -= d[i2] ^ (d[i3] << 10 | d[i3] >> 22); + i1--; i2--; i3--; + + d[i1] -= d[i2] ^ d[i3]; + i1--; i2--; i3--; + } + } +} + +static int crypt_iv_elephant(struct crypt_config *cc, struct dm_crypt_request *dmreq) +{ + struct iv_elephant_private *elephant = &cc->iv_gen_private.elephant; + u8 *es, *ks, *data, *data2, *data_offset; + struct skcipher_request *req; + struct scatterlist *sg, *sg2, src, dst; + struct crypto_wait wait; + int i, r; + + req = skcipher_request_alloc(elephant->tfm, GFP_NOIO); + es = kzalloc(16, GFP_NOIO); /* Key for AES */ + ks = kzalloc(32, GFP_NOIO); /* Elephant sector key */ + + if (!req || !es || !ks) { + r = -ENOMEM; + goto out; + } + + *(__le64 *)es = cpu_to_le64(dmreq->iv_sector * cc->sector_size); + + /* E(Ks, e(s)) */ + sg_init_one(&src, es, 16); + sg_init_one(&dst, ks, 16); + skcipher_request_set_crypt(req, &src, &dst, 16, NULL); + skcipher_request_set_callback(req, 0, crypto_req_done, &wait); + r = crypto_wait_req(crypto_skcipher_encrypt(req), &wait); + if (r) + goto out; + + /* E(Ks, e'(s)) */ + es[15] = 0x80; + sg_init_one(&dst, &ks[16], 16); + r = crypto_wait_req(crypto_skcipher_encrypt(req), &wait); + if (r) + goto out; + + sg = crypt_get_sg_data(cc, dmreq->sg_out); + data = kmap_atomic(sg_page(sg)); + data_offset = data + sg->offset; + + /* Cannot modify original bio, copy to sg_out and apply Elephant to it */ + if (bio_data_dir(dmreq->ctx->bio_in) == WRITE) { + sg2 = crypt_get_sg_data(cc, dmreq->sg_in); + data2 = kmap_atomic(sg_page(sg2)); + memcpy(data_offset, data2 + sg2->offset, cc->sector_size); + kunmap_atomic(data2); + } + + if (bio_data_dir(dmreq->ctx->bio_in) != WRITE) { + diffuser_disk_to_cpu((u32*)data_offset, cc->sector_size / sizeof(u32)); + diffuser_b_decrypt((u32*)data_offset, cc->sector_size / sizeof(u32)); + diffuser_a_decrypt((u32*)data_offset, cc->sector_size / sizeof(u32)); + diffuser_cpu_to_disk((__le32*)data_offset, cc->sector_size / sizeof(u32)); + } + + for (i = 0; i < (cc->sector_size / 32); i++) + crypto_xor(data_offset + i * 32, ks, 32); + + if (bio_data_dir(dmreq->ctx->bio_in) == WRITE) { + diffuser_disk_to_cpu((u32*)data_offset, cc->sector_size / sizeof(u32)); + diffuser_a_encrypt((u32*)data_offset, cc->sector_size / sizeof(u32)); + diffuser_b_encrypt((u32*)data_offset, cc->sector_size / sizeof(u32)); + diffuser_cpu_to_disk((__le32*)data_offset, cc->sector_size / sizeof(u32)); + } + + kunmap_atomic(data); +out: + kzfree(ks); + kzfree(es); + skcipher_request_free(req); + return r; +} + +static int crypt_iv_elephant_gen(struct crypt_config *cc, u8 *iv, + struct dm_crypt_request *dmreq) +{ + int r; + + if (bio_data_dir(dmreq->ctx->bio_in) == WRITE) { + r = crypt_iv_elephant(cc, dmreq); + if (r) + return r; + } + + return crypt_iv_eboiv_gen(cc, iv, dmreq); +} + +static int crypt_iv_elephant_post(struct crypt_config *cc, u8 *iv, + struct dm_crypt_request *dmreq) +{ + if (bio_data_dir(dmreq->ctx->bio_in) != WRITE) + return crypt_iv_elephant(cc, dmreq); + + return 0; +} + +static int crypt_iv_elephant_init(struct crypt_config *cc) +{ + struct iv_elephant_private *elephant = &cc->iv_gen_private.elephant; + int key_offset = cc->key_size - cc->key_extra_size; + + return crypto_skcipher_setkey(elephant->tfm, &cc->key[key_offset], cc->key_extra_size); +} + +static int crypt_iv_elephant_wipe(struct crypt_config *cc) +{ + struct iv_elephant_private *elephant = &cc->iv_gen_private.elephant; + u8 key[ELEPHANT_MAX_KEY_SIZE]; + + memset(key, 0, cc->key_extra_size); + return crypto_skcipher_setkey(elephant->tfm, key, cc->key_extra_size); +} + static const struct crypt_iv_operations crypt_iv_plain_ops = { .generator = crypt_iv_plain_gen }; @@ -787,6 +1089,15 @@ static struct crypt_iv_operations crypt_iv_eboiv_ops = { .generator = crypt_iv_eboiv_gen }; +static struct crypt_iv_operations crypt_iv_elephant_ops = { + .ctr = crypt_iv_elephant_ctr, + .dtr = crypt_iv_elephant_dtr, + .init = crypt_iv_elephant_init, + .wipe = crypt_iv_elephant_wipe, + .generator = crypt_iv_elephant_gen, + .post = crypt_iv_elephant_post +}; + /* * Integrity extensions */ @@ -1103,6 +1414,9 @@ static int crypt_convert_block_skcipher(struct crypt_config *cc, r = cc->iv_gen_ops->generator(cc, org_iv, dmreq); if (r < 0) return r; + /* Data can be already preprocessed in generator */ + if (test_bit(CRYPT_ENCRYPT_PREPROCESS, &cc->cipher_flags)) + sg_in = sg_out; /* Store generated IV in integrity metadata */ if (cc->integrity_iv_size) memcpy(tag_iv, org_iv, cc->integrity_iv_size); @@ -2191,7 +2505,14 @@ static int crypt_ctr_ivmode(struct dm_target *ti, const char *ivmode) cc->iv_gen_ops = &crypt_iv_null_ops; else if (strcmp(ivmode, "eboiv") == 0) cc->iv_gen_ops = &crypt_iv_eboiv_ops; - else if (strcmp(ivmode, "lmk") == 0) { + else if (strcmp(ivmode, "elephant") == 0) { + cc->iv_gen_ops = &crypt_iv_elephant_ops; + cc->key_parts = 2; + cc->key_extra_size = cc->key_size / 2; + if (cc->key_extra_size > ELEPHANT_MAX_KEY_SIZE) + return -EINVAL; + set_bit(CRYPT_ENCRYPT_PREPROCESS, &cc->cipher_flags); + } else if (strcmp(ivmode, "lmk") == 0) { cc->iv_gen_ops = &crypt_iv_lmk_ops; /* * Version 2 and 3 is recognised according @@ -2959,7 +3280,7 @@ static void crypt_io_hints(struct dm_target *ti, struct queue_limits *limits) static struct target_type crypt_target = { .name = "crypt", - .version = {1, 19, 0}, + .version = {1, 20, 0}, .module = THIS_MODULE, .ctr = crypt_ctr, .dtr = crypt_dtr, diff --git a/drivers/md/dm-dust.c b/drivers/md/dm-dust.c index eb37584427a4..ff03b90072c5 100644 --- a/drivers/md/dm-dust.c +++ b/drivers/md/dm-dust.c @@ -207,16 +207,16 @@ static int dust_map_write(struct dust_device *dd, sector_t thisblock, bool fail_read_on_bb) { unsigned long flags; - int ret = DM_MAPIO_REMAPPED; + int r = DM_MAPIO_REMAPPED; if (fail_read_on_bb) { thisblock >>= dd->sect_per_block_shift; spin_lock_irqsave(&dd->dust_lock, flags); - ret = __dust_map_write(dd, thisblock); + r = __dust_map_write(dd, thisblock); spin_unlock_irqrestore(&dd->dust_lock, flags); } - return ret; + return r; } static int dust_map(struct dm_target *ti, struct bio *bio) diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c index e0c32793c248..2bc18c9c3abc 100644 --- a/drivers/md/dm-mpath.c +++ b/drivers/md/dm-mpath.c @@ -20,6 +20,7 @@ #include <linux/pagemap.h> #include <linux/slab.h> #include <linux/time.h> +#include <linux/timer.h> #include <linux/workqueue.h> #include <linux/delay.h> #include <scsi/scsi_dh.h> @@ -29,6 +30,9 @@ #define DM_MSG_PREFIX "multipath" #define DM_PG_INIT_DELAY_MSECS 2000 #define DM_PG_INIT_DELAY_DEFAULT ((unsigned) -1) +#define QUEUE_IF_NO_PATH_TIMEOUT_DEFAULT 0 + +static unsigned long queue_if_no_path_timeout_secs = QUEUE_IF_NO_PATH_TIMEOUT_DEFAULT; /* Path properties */ struct pgpath { @@ -91,6 +95,8 @@ struct multipath { struct work_struct process_queued_bios; struct bio_list queued_bios; + + struct timer_list nopath_timer; /* Timeout for queue_if_no_path */ }; /* @@ -108,6 +114,7 @@ static void trigger_event(struct work_struct *work); static void activate_or_offline_path(struct pgpath *pgpath); static void activate_path_work(struct work_struct *work); static void process_queued_bios(struct work_struct *work); +static void queue_if_no_path_timeout_work(struct timer_list *t); /*----------------------------------------------- * Multipath state flags. @@ -195,6 +202,8 @@ static struct multipath *alloc_multipath(struct dm_target *ti) m->ti = ti; ti->private = m; + + timer_setup(&m->nopath_timer, queue_if_no_path_timeout_work, 0); } return m; @@ -718,6 +727,43 @@ static int queue_if_no_path(struct multipath *m, bool queue_if_no_path, } /* + * If the queue_if_no_path timeout fires, turn off queue_if_no_path and + * process any queued I/O. + */ +static void queue_if_no_path_timeout_work(struct timer_list *t) +{ + struct multipath *m = from_timer(m, t, nopath_timer); + struct mapped_device *md = dm_table_get_md(m->ti->table); + + DMWARN("queue_if_no_path timeout on %s, failing queued IO", dm_device_name(md)); + queue_if_no_path(m, false, false); +} + +/* + * Enable the queue_if_no_path timeout if necessary. + * Called with m->lock held. + */ +static void enable_nopath_timeout(struct multipath *m) +{ + unsigned long queue_if_no_path_timeout = + READ_ONCE(queue_if_no_path_timeout_secs) * HZ; + + lockdep_assert_held(&m->lock); + + if (queue_if_no_path_timeout > 0 && + atomic_read(&m->nr_valid_paths) == 0 && + test_bit(MPATHF_QUEUE_IF_NO_PATH, &m->flags)) { + mod_timer(&m->nopath_timer, + jiffies + queue_if_no_path_timeout); + } +} + +static void disable_nopath_timeout(struct multipath *m) +{ + del_timer_sync(&m->nopath_timer); +} + +/* * An event is triggered whenever a path is taken out of use. * Includes path failure and PG bypass. */ @@ -1090,6 +1136,7 @@ static int multipath_ctr(struct dm_target *ti, unsigned argc, char **argv) struct dm_arg_set as; unsigned pg_count = 0; unsigned next_pg_num; + unsigned long flags; as.argc = argc; as.argv = argv; @@ -1154,6 +1201,10 @@ static int multipath_ctr(struct dm_target *ti, unsigned argc, char **argv) goto bad; } + spin_lock_irqsave(&m->lock, flags); + enable_nopath_timeout(m); + spin_unlock_irqrestore(&m->lock, flags); + ti->num_flush_bios = 1; ti->num_discard_bios = 1; ti->num_write_same_bios = 1; @@ -1208,6 +1259,7 @@ static void multipath_dtr(struct dm_target *ti) { struct multipath *m = ti->private; + disable_nopath_timeout(m); flush_multipath_work(m); free_multipath(m); } @@ -1241,6 +1293,8 @@ static int fail_path(struct pgpath *pgpath) schedule_work(&m->trigger_event); + enable_nopath_timeout(m); + out: spin_unlock_irqrestore(&m->lock, flags); @@ -1291,6 +1345,9 @@ out: process_queued_io_list(m); } + if (pgpath->is_active) + disable_nopath_timeout(m); + return r; } @@ -1444,7 +1501,7 @@ static void pg_init_done(void *data, int errors) break; case SCSI_DH_RETRY: /* Wait before retrying. */ - delay_retry = 1; + delay_retry = true; /* fall through */ case SCSI_DH_IMM_RETRY: case SCSI_DH_RES_TEMP_UNAVAIL: @@ -1789,6 +1846,7 @@ static int multipath_message(struct dm_target *ti, unsigned argc, char **argv, struct dm_dev *dev; struct multipath *m = ti->private; action_fn action; + unsigned long flags; mutex_lock(&m->work_mutex); @@ -1800,9 +1858,13 @@ static int multipath_message(struct dm_target *ti, unsigned argc, char **argv, if (argc == 1) { if (!strcasecmp(argv[0], "queue_if_no_path")) { r = queue_if_no_path(m, true, false); + spin_lock_irqsave(&m->lock, flags); + enable_nopath_timeout(m); + spin_unlock_irqrestore(&m->lock, flags); goto out; } else if (!strcasecmp(argv[0], "fail_if_no_path")) { r = queue_if_no_path(m, false, false); + disable_nopath_timeout(m); goto out; } } @@ -2065,6 +2127,10 @@ static void __exit dm_multipath_exit(void) module_init(dm_multipath_init); module_exit(dm_multipath_exit); +module_param_named(queue_if_no_path_timeout_secs, + queue_if_no_path_timeout_secs, ulong, S_IRUGO | S_IWUSR); +MODULE_PARM_DESC(queue_if_no_path_timeout_secs, "No available paths queue IO timeout in seconds"); + MODULE_DESCRIPTION(DM_NAME " multipath target"); MODULE_AUTHOR("Sistina Software <dm-devel@redhat.com>"); MODULE_LICENSE("GPL"); diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c index c412eaa975fc..9a18bef0a5ff 100644 --- a/drivers/md/dm-raid.c +++ b/drivers/md/dm-raid.c @@ -129,7 +129,9 @@ struct raid_dev { CTR_FLAG_RAID10_COPIES | \ CTR_FLAG_RAID10_FORMAT | \ CTR_FLAG_DELTA_DISKS | \ - CTR_FLAG_DATA_OFFSET) + CTR_FLAG_DATA_OFFSET | \ + CTR_FLAG_JOURNAL_DEV | \ + CTR_FLAG_JOURNAL_MODE) /* Valid options definitions per raid level... */ @@ -3001,7 +3003,6 @@ static int raid_ctr(struct dm_target *ti, unsigned int argc, char **argv) { 1, 254, "Cannot understand number of raid devices parameters" } }; - /* Must have <raid_type> */ arg = dm_shift_arg(&as); if (!arg) { ti->error = "No arguments"; @@ -3508,8 +3509,7 @@ static void raid_status(struct dm_target *ti, status_type_t type, unsigned long recovery; 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; + unsigned int rebuild_writemostly_count = 0; sector_t progress, resync_max_sectors, resync_mismatches; enum sync_state state; struct raid_type *rt; @@ -3593,18 +3593,20 @@ static void raid_status(struct dm_target *ti, status_type_t type, 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 + + /* + * Count any rebuild or writemostly argument pairs and subtract the + * hweight count being added below of any rebuild and writemostly ctr flags. + */ + for (i = 0; i < rs->raid_disks; i++) { + rebuild_writemostly_count += (test_bit(i, (void *) rs->rebuild_disks) ? 2 : 0) + + (test_bit(WriteMostly, &rs->dev[i].rdev.flags) ? 2 : 0); + } + rebuild_writemostly_count -= (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags) ? 2 : 0) + + (test_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags) ? 2 : 0); + /* Calculate raid parameter count based on ^ rebuild/writemostly argument counts and ctr flags set. */ + raid_param_cnt += rebuild_writemostly_count + hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_NO_ARGS) + - hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_ONE_ARG) * 2 + - (test_bit(__CTR_FLAG_JOURNAL_DEV, &rs->ctr_flags) ? 2 : 0) + - (test_bit(__CTR_FLAG_JOURNAL_MODE, &rs->ctr_flags) ? 2 : 0); - + hweight32(rs->ctr_flags & CTR_FLAG_OPTIONS_ONE_ARG) * 2; /* Emit table line */ /* This has to be in the documented order for userspace! */ DMEMIT("%s %u %u", rs->raid_type->name, raid_param_cnt, mddev->new_chunk_sectors); @@ -3612,11 +3614,10 @@ static void raid_status(struct dm_target *ti, status_type_t type, DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_SYNC)); if (test_bit(__CTR_FLAG_NOSYNC, &rs->ctr_flags)) DMEMIT(" %s", dm_raid_arg_name_by_flag(CTR_FLAG_NOSYNC)); - if (rebuild_disks) + if (test_bit(__CTR_FLAG_REBUILD, &rs->ctr_flags)) 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 (test_bit(i, (void *) rs->rebuild_disks)) + DMEMIT(" %s %u", dm_raid_arg_name_by_flag(CTR_FLAG_REBUILD), i); 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); @@ -3626,7 +3627,7 @@ static void raid_status(struct dm_target *ti, status_type_t type, 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 (write_mostly_params) + if (test_bit(__CTR_FLAG_WRITE_MOSTLY, &rs->ctr_flags)) 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), @@ -4029,7 +4030,7 @@ static void raid_resume(struct dm_target *ti) static struct target_type raid_target = { .name = "raid", - .version = {1, 15, 0}, + .version = {1, 15, 1}, .module = THIS_MODULE, .ctr = raid_ctr, .dtr = raid_dtr, diff --git a/drivers/md/dm-snap.c b/drivers/md/dm-snap.c index 4fb1a40e68a0..6b11a266299f 100644 --- a/drivers/md/dm-snap.c +++ b/drivers/md/dm-snap.c @@ -1061,7 +1061,7 @@ static void snapshot_merge_next_chunks(struct dm_snapshot *s) DMERR("Read error in exception store: " "shutting down merge"); down_write(&s->lock); - s->merge_failed = 1; + s->merge_failed = true; up_write(&s->lock); } goto shut; @@ -1149,7 +1149,7 @@ static void merge_callback(int read_err, unsigned long write_err, void *context) shut: down_write(&s->lock); - s->merge_failed = 1; + s->merge_failed = true; b = __release_queued_bios_after_merge(s); up_write(&s->lock); error_bios(b); @@ -1314,7 +1314,7 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv) INIT_LIST_HEAD(&s->list); spin_lock_init(&s->pe_lock); s->state_bits = 0; - s->merge_failed = 0; + s->merge_failed = false; s->first_merging_chunk = 0; s->num_merging_chunks = 0; bio_list_init(&s->bios_queued_during_merge); diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c index b88d6d701f5b..fc9947d6210c 100644 --- a/drivers/md/dm-thin-metadata.c +++ b/drivers/md/dm-thin-metadata.c @@ -28,7 +28,7 @@ * * - A hierarchical btree, with 2 levels which effectively maps (thin * dev id, virtual block) -> block_time. Block time is a 64-bit - * field holding the time in the low 24 bits, and block in the top 48 + * field holding the time in the low 24 bits, and block in the top 40 * bits. * * BTrees consist solely of btree_nodes, that fill a block. Some are @@ -387,16 +387,15 @@ static int subtree_equal(void *context, const void *value1_le, const void *value * Variant that is used for in-core only changes or code that * shouldn't put the pool in service on its own (e.g. commit). */ -static inline void __pmd_write_lock(struct dm_pool_metadata *pmd) +static inline void pmd_write_lock_in_core(struct dm_pool_metadata *pmd) __acquires(pmd->root_lock) { down_write(&pmd->root_lock); } -#define pmd_write_lock_in_core(pmd) __pmd_write_lock((pmd)) static inline void pmd_write_lock(struct dm_pool_metadata *pmd) { - __pmd_write_lock(pmd); + pmd_write_lock_in_core(pmd); if (unlikely(!pmd->in_service)) pmd->in_service = true; } @@ -811,7 +810,7 @@ static int __write_changed_details(struct dm_pool_metadata *pmd) return r; if (td->open_count) - td->changed = 0; + td->changed = false; else { list_del(&td->list); kfree(td); @@ -831,6 +830,7 @@ static int __commit_transaction(struct dm_pool_metadata *pmd) * We need to know if the thin_disk_superblock exceeds a 512-byte sector. */ BUILD_BUG_ON(sizeof(struct thin_disk_superblock) > 512); + BUG_ON(!rwsem_is_locked(&pmd->root_lock)); if (unlikely(!pmd->in_service)) return 0; @@ -953,6 +953,7 @@ int dm_pool_metadata_close(struct dm_pool_metadata *pmd) return -EBUSY; } + pmd_write_lock_in_core(pmd); if (!dm_bm_is_read_only(pmd->bm) && !pmd->fail_io) { r = __commit_transaction(pmd); if (r < 0) @@ -961,6 +962,7 @@ int dm_pool_metadata_close(struct dm_pool_metadata *pmd) } if (!pmd->fail_io) __destroy_persistent_data_objects(pmd); + pmd_write_unlock(pmd); kfree(pmd); return 0; @@ -1106,7 +1108,7 @@ static int __set_snapshot_details(struct dm_pool_metadata *pmd, if (r) return r; - td->changed = 1; + td->changed = true; td->snapshotted_time = time; snap->mapped_blocks = td->mapped_blocks; @@ -1618,7 +1620,7 @@ static int __insert(struct dm_thin_device *td, dm_block_t block, if (r) return r; - td->changed = 1; + td->changed = true; if (inserted) td->mapped_blocks++; @@ -1649,7 +1651,7 @@ static int __remove(struct dm_thin_device *td, dm_block_t block) return r; td->mapped_blocks--; - td->changed = 1; + td->changed = true; return 0; } @@ -1703,7 +1705,7 @@ static int __remove_range(struct dm_thin_device *td, dm_block_t begin, dm_block_ } td->mapped_blocks -= total_count; - td->changed = 1; + td->changed = true; /* * Reinsert the mapping tree. @@ -1841,7 +1843,7 @@ int dm_pool_commit_metadata(struct dm_pool_metadata *pmd) * Care is taken to not have commit be what * triggers putting the thin-pool in-service. */ - __pmd_write_lock(pmd); + pmd_write_lock_in_core(pmd); if (pmd->fail_io) goto out; diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c index 57626c27a54b..fa8d5464c1fb 100644 --- a/drivers/md/dm-thin.c +++ b/drivers/md/dm-thin.c @@ -231,6 +231,7 @@ struct pool { struct dm_target *ti; /* Only set if a pool target is bound */ struct mapped_device *pool_md; + struct block_device *data_dev; struct block_device *md_dev; struct dm_pool_metadata *pmd; @@ -281,6 +282,8 @@ struct pool { struct dm_bio_prison_cell **cell_sort_array; mempool_t mapping_pool; + + struct bio flush_bio; }; static void metadata_operation_failed(struct pool *pool, const char *op, int r); @@ -328,7 +331,6 @@ struct pool_c { dm_block_t low_water_blocks; struct pool_features requested_pf; /* Features requested during table load */ struct pool_features adjusted_pf; /* Features used after adjusting for constituent devices */ - struct bio flush_bio; }; /* @@ -2924,6 +2926,7 @@ static void __pool_destroy(struct pool *pool) if (pool->next_mapping) mempool_free(pool->next_mapping, &pool->mapping_pool); mempool_exit(&pool->mapping_pool); + bio_uninit(&pool->flush_bio); dm_deferred_set_destroy(pool->shared_read_ds); dm_deferred_set_destroy(pool->all_io_ds); kfree(pool); @@ -2933,6 +2936,7 @@ static struct kmem_cache *_new_mapping_cache; static struct pool *pool_create(struct mapped_device *pool_md, struct block_device *metadata_dev, + struct block_device *data_dev, unsigned long block_size, int read_only, char **error) { @@ -3003,6 +3007,7 @@ static struct pool *pool_create(struct mapped_device *pool_md, pool->low_water_triggered = false; pool->suspended = true; pool->out_of_data_space = false; + bio_init(&pool->flush_bio, NULL, 0); pool->shared_read_ds = dm_deferred_set_create(); if (!pool->shared_read_ds) { @@ -3040,6 +3045,7 @@ static struct pool *pool_create(struct mapped_device *pool_md, pool->last_commit_jiffies = jiffies; pool->pool_md = pool_md; pool->md_dev = metadata_dev; + pool->data_dev = data_dev; __pool_table_insert(pool); return pool; @@ -3081,6 +3087,7 @@ static void __pool_dec(struct pool *pool) static struct pool *__pool_find(struct mapped_device *pool_md, struct block_device *metadata_dev, + struct block_device *data_dev, unsigned long block_size, int read_only, char **error, int *created) { @@ -3091,19 +3098,23 @@ static struct pool *__pool_find(struct mapped_device *pool_md, *error = "metadata device already in use by a pool"; return ERR_PTR(-EBUSY); } + if (pool->data_dev != data_dev) { + *error = "data device already in use by a pool"; + return ERR_PTR(-EBUSY); + } __pool_inc(pool); } else { pool = __pool_table_lookup(pool_md); if (pool) { - if (pool->md_dev != metadata_dev) { + if (pool->md_dev != metadata_dev || pool->data_dev != data_dev) { *error = "different pool cannot replace a pool"; return ERR_PTR(-EINVAL); } __pool_inc(pool); } else { - pool = pool_create(pool_md, metadata_dev, block_size, read_only, error); + pool = pool_create(pool_md, metadata_dev, data_dev, block_size, read_only, error); *created = 1; } } @@ -3124,7 +3135,6 @@ static void pool_dtr(struct dm_target *ti) __pool_dec(pt->pool); dm_put_device(ti, pt->metadata_dev); dm_put_device(ti, pt->data_dev); - bio_uninit(&pt->flush_bio); kfree(pt); mutex_unlock(&dm_thin_pool_table.mutex); @@ -3203,11 +3213,11 @@ static void metadata_low_callback(void *context) */ static int metadata_pre_commit_callback(void *context) { - struct pool_c *pt = context; - struct bio *flush_bio = &pt->flush_bio; + struct pool *pool = context; + struct bio *flush_bio = &pool->flush_bio; bio_reset(flush_bio); - bio_set_dev(flush_bio, pt->data_dev->bdev); + bio_set_dev(flush_bio, pool->data_dev); flush_bio->bi_opf = REQ_OP_WRITE | REQ_PREFLUSH; return submit_bio_wait(flush_bio); @@ -3356,7 +3366,7 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv) goto out; } - pool = __pool_find(dm_table_get_md(ti->table), metadata_dev->bdev, + pool = __pool_find(dm_table_get_md(ti->table), metadata_dev->bdev, data_dev->bdev, block_size, pf.mode == PM_READ_ONLY, &ti->error, &pool_created); if (IS_ERR(pool)) { r = PTR_ERR(pool); @@ -3381,7 +3391,6 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv) pt->data_dev = data_dev; pt->low_water_blocks = low_water_blocks; pt->adjusted_pf = pt->requested_pf = pf; - bio_init(&pt->flush_bio, NULL, 0); ti->num_flush_bios = 1; /* @@ -3408,9 +3417,8 @@ static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv) if (r) goto out_flags_changed; - dm_pool_register_pre_commit_callback(pt->pool->pmd, - metadata_pre_commit_callback, - pt); + dm_pool_register_pre_commit_callback(pool->pmd, + metadata_pre_commit_callback, pool); pt->callbacks.congested_fn = pool_is_congested; dm_table_add_target_callbacks(ti->table, &pt->callbacks); @@ -4099,7 +4107,7 @@ static struct target_type pool_target = { .name = "thin-pool", .features = DM_TARGET_SINGLETON | DM_TARGET_ALWAYS_WRITEABLE | DM_TARGET_IMMUTABLE, - .version = {1, 21, 0}, + .version = {1, 22, 0}, .module = THIS_MODULE, .ctr = pool_ctr, .dtr = pool_dtr, @@ -4476,7 +4484,7 @@ static void thin_io_hints(struct dm_target *ti, struct queue_limits *limits) static struct target_type thin_target = { .name = "thin", - .version = {1, 21, 0}, + .version = {1, 22, 0}, .module = THIS_MODULE, .ctr = thin_ctr, .dtr = thin_dtr, diff --git a/drivers/md/dm-verity-target.c b/drivers/md/dm-verity-target.c index 4fb33e7562c5..0d61e9c67986 100644 --- a/drivers/md/dm-verity-target.c +++ b/drivers/md/dm-verity-target.c @@ -611,8 +611,22 @@ no_prefetch_cluster: static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io) { + sector_t block = io->block; + unsigned int n_blocks = io->n_blocks; struct dm_verity_prefetch_work *pw; + if (v->validated_blocks) { + while (n_blocks && test_bit(block, v->validated_blocks)) { + block++; + n_blocks--; + } + while (n_blocks && test_bit(block + n_blocks - 1, + v->validated_blocks)) + n_blocks--; + if (!n_blocks) + return; + } + pw = kmalloc(sizeof(struct dm_verity_prefetch_work), GFP_NOIO | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN); @@ -621,8 +635,8 @@ static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io) INIT_WORK(&pw->work, verity_prefetch_io); pw->v = v; - pw->block = io->block; - pw->n_blocks = io->n_blocks; + pw->block = block; + pw->n_blocks = n_blocks; queue_work(v->verify_wq, &pw->work); } diff --git a/drivers/md/dm-writecache.c b/drivers/md/dm-writecache.c index 7d727a72aa13..b9e27e37a943 100644 --- a/drivers/md/dm-writecache.c +++ b/drivers/md/dm-writecache.c @@ -442,7 +442,13 @@ static void writecache_notify_io(unsigned long error, void *context) complete(&endio->c); } -static void ssd_commit_flushed(struct dm_writecache *wc) +static void writecache_wait_for_ios(struct dm_writecache *wc, int direction) +{ + wait_event(wc->bio_in_progress_wait[direction], + !atomic_read(&wc->bio_in_progress[direction])); +} + +static void ssd_commit_flushed(struct dm_writecache *wc, bool wait_for_ios) { struct dm_io_region region; struct dm_io_request req; @@ -488,17 +494,20 @@ static void ssd_commit_flushed(struct dm_writecache *wc) writecache_notify_io(0, &endio); wait_for_completion_io(&endio.c); + if (wait_for_ios) + writecache_wait_for_ios(wc, WRITE); + writecache_disk_flush(wc, wc->ssd_dev); memset(wc->dirty_bitmap, 0, wc->dirty_bitmap_size); } -static void writecache_commit_flushed(struct dm_writecache *wc) +static void writecache_commit_flushed(struct dm_writecache *wc, bool wait_for_ios) { if (WC_MODE_PMEM(wc)) wmb(); else - ssd_commit_flushed(wc); + ssd_commit_flushed(wc, wait_for_ios); } static void writecache_disk_flush(struct dm_writecache *wc, struct dm_dev *dev) @@ -522,12 +531,6 @@ static void writecache_disk_flush(struct dm_writecache *wc, struct dm_dev *dev) writecache_error(wc, r, "error flushing metadata: %d", r); } -static void writecache_wait_for_ios(struct dm_writecache *wc, int direction) -{ - wait_event(wc->bio_in_progress_wait[direction], - !atomic_read(&wc->bio_in_progress[direction])); -} - #define WFE_RETURN_FOLLOWING 1 #define WFE_LOWEST_SEQ 2 @@ -622,7 +625,7 @@ static void writecache_add_to_freelist(struct dm_writecache *wc, struct wc_entry wc->freelist_size++; } -static struct wc_entry *writecache_pop_from_freelist(struct dm_writecache *wc) +static struct wc_entry *writecache_pop_from_freelist(struct dm_writecache *wc, sector_t expected_sector) { struct wc_entry *e; @@ -631,6 +634,8 @@ static struct wc_entry *writecache_pop_from_freelist(struct dm_writecache *wc) if (unlikely(!wc->current_free)) return NULL; e = wc->current_free; + if (expected_sector != (sector_t)-1 && unlikely(cache_sector(wc, e) != expected_sector)) + return NULL; next = rb_next(&e->rb_node); rb_erase(&e->rb_node, &wc->freetree); if (unlikely(!next)) @@ -640,6 +645,8 @@ static struct wc_entry *writecache_pop_from_freelist(struct dm_writecache *wc) if (unlikely(list_empty(&wc->freelist))) return NULL; e = container_of(wc->freelist.next, struct wc_entry, lru); + if (expected_sector != (sector_t)-1 && unlikely(cache_sector(wc, e) != expected_sector)) + return NULL; list_del(&e->lru); } wc->freelist_size--; @@ -724,15 +731,12 @@ static void writecache_flush(struct dm_writecache *wc) e = e2; cond_resched(); } - writecache_commit_flushed(wc); - - if (!WC_MODE_PMEM(wc)) - writecache_wait_for_ios(wc, WRITE); + writecache_commit_flushed(wc, true); wc->seq_count++; pmem_assign(sb(wc)->seq_count, cpu_to_le64(wc->seq_count)); writecache_flush_region(wc, &sb(wc)->seq_count, sizeof sb(wc)->seq_count); - writecache_commit_flushed(wc); + writecache_commit_flushed(wc, false); wc->overwrote_committed = false; @@ -756,7 +760,7 @@ static void writecache_flush(struct dm_writecache *wc) } if (need_flush_after_free) - writecache_commit_flushed(wc); + writecache_commit_flushed(wc, false); } static void writecache_flush_work(struct work_struct *work) @@ -809,7 +813,7 @@ static void writecache_discard(struct dm_writecache *wc, sector_t start, sector_ } if (discarded_something) - writecache_commit_flushed(wc); + writecache_commit_flushed(wc, false); } static bool writecache_wait_for_writeback(struct dm_writecache *wc) @@ -958,7 +962,7 @@ erase_this: if (need_flush) { writecache_flush_all_metadata(wc); - writecache_commit_flushed(wc); + writecache_commit_flushed(wc, false); } wc_unlock(wc); @@ -1193,7 +1197,7 @@ read_next_block: goto bio_copy; } } - e = writecache_pop_from_freelist(wc); + e = writecache_pop_from_freelist(wc, (sector_t)-1); if (unlikely(!e)) { writecache_wait_on_freelist(wc); continue; @@ -1205,9 +1209,26 @@ bio_copy: if (WC_MODE_PMEM(wc)) { bio_copy_block(wc, bio, memory_data(wc, e)); } else { - dm_accept_partial_bio(bio, wc->block_size >> SECTOR_SHIFT); + unsigned bio_size = wc->block_size; + sector_t start_cache_sec = cache_sector(wc, e); + sector_t current_cache_sec = start_cache_sec + (bio_size >> SECTOR_SHIFT); + + while (bio_size < bio->bi_iter.bi_size) { + struct wc_entry *f = writecache_pop_from_freelist(wc, current_cache_sec); + if (!f) + break; + write_original_sector_seq_count(wc, f, bio->bi_iter.bi_sector + + (bio_size >> SECTOR_SHIFT), wc->seq_count); + writecache_insert_entry(wc, f); + wc->uncommitted_blocks++; + bio_size += wc->block_size; + current_cache_sec += wc->block_size >> SECTOR_SHIFT; + } + bio_set_dev(bio, wc->ssd_dev->bdev); - bio->bi_iter.bi_sector = cache_sector(wc, e); + bio->bi_iter.bi_sector = start_cache_sec; + dm_accept_partial_bio(bio, bio_size >> SECTOR_SHIFT); + if (unlikely(wc->uncommitted_blocks >= wc->autocommit_blocks)) { wc->uncommitted_blocks = 0; queue_work(wc->writeback_wq, &wc->flush_work); @@ -1342,7 +1363,7 @@ static void __writecache_endio_pmem(struct dm_writecache *wc, struct list_head * wc->writeback_size--; n_walked++; if (unlikely(n_walked >= ENDIO_LATENCY)) { - writecache_commit_flushed(wc); + writecache_commit_flushed(wc, false); wc_unlock(wc); wc_lock(wc); n_walked = 0; @@ -1423,7 +1444,7 @@ pop_from_list: writecache_wait_for_ios(wc, READ); } - writecache_commit_flushed(wc); + writecache_commit_flushed(wc, false); wc_unlock(wc); } @@ -1766,10 +1787,10 @@ static int init_memory(struct dm_writecache *wc) write_original_sector_seq_count(wc, &wc->entries[b], -1, -1); writecache_flush_all_metadata(wc); - writecache_commit_flushed(wc); + writecache_commit_flushed(wc, false); pmem_assign(sb(wc)->magic, cpu_to_le32(MEMORY_SUPERBLOCK_MAGIC)); writecache_flush_region(wc, &sb(wc)->magic, sizeof sb(wc)->magic); - writecache_commit_flushed(wc); + writecache_commit_flushed(wc, false); return 0; } diff --git a/drivers/md/dm-zoned-metadata.c b/drivers/md/dm-zoned-metadata.c index 22b3cb0050a7..516c7b671d25 100644 --- a/drivers/md/dm-zoned-metadata.c +++ b/drivers/md/dm-zoned-metadata.c @@ -134,6 +134,7 @@ struct dmz_metadata { sector_t zone_bitmap_size; unsigned int zone_nr_bitmap_blocks; + unsigned int zone_bits_per_mblk; unsigned int nr_bitmap_blocks; unsigned int nr_map_blocks; @@ -1161,7 +1162,10 @@ static int dmz_init_zones(struct dmz_metadata *zmd) /* Init */ zmd->zone_bitmap_size = dev->zone_nr_blocks >> 3; - zmd->zone_nr_bitmap_blocks = zmd->zone_bitmap_size >> DMZ_BLOCK_SHIFT; + zmd->zone_nr_bitmap_blocks = + max_t(sector_t, 1, zmd->zone_bitmap_size >> DMZ_BLOCK_SHIFT); + zmd->zone_bits_per_mblk = min_t(sector_t, dev->zone_nr_blocks, + DMZ_BLOCK_SIZE_BITS); /* Allocate zone array */ zmd->zones = kcalloc(dev->nr_zones, sizeof(struct dm_zone), GFP_KERNEL); @@ -1956,7 +1960,7 @@ int dmz_copy_valid_blocks(struct dmz_metadata *zmd, struct dm_zone *from_zone, dmz_release_mblock(zmd, to_mblk); dmz_release_mblock(zmd, from_mblk); - chunk_block += DMZ_BLOCK_SIZE_BITS; + chunk_block += zmd->zone_bits_per_mblk; } to_zone->weight = from_zone->weight; @@ -2017,7 +2021,7 @@ int dmz_validate_blocks(struct dmz_metadata *zmd, struct dm_zone *zone, /* Set bits */ bit = chunk_block & DMZ_BLOCK_MASK_BITS; - nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit); + nr_bits = min(nr_blocks, zmd->zone_bits_per_mblk - bit); count = dmz_set_bits((unsigned long *)mblk->data, bit, nr_bits); if (count) { @@ -2096,7 +2100,7 @@ int dmz_invalidate_blocks(struct dmz_metadata *zmd, struct dm_zone *zone, /* Clear bits */ bit = chunk_block & DMZ_BLOCK_MASK_BITS; - nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit); + nr_bits = min(nr_blocks, zmd->zone_bits_per_mblk - bit); count = dmz_clear_bits((unsigned long *)mblk->data, bit, nr_bits); @@ -2156,6 +2160,7 @@ static int dmz_to_next_set_block(struct dmz_metadata *zmd, struct dm_zone *zone, { struct dmz_mblock *mblk; unsigned int bit, set_bit, nr_bits; + unsigned int zone_bits = zmd->zone_bits_per_mblk; unsigned long *bitmap; int n = 0; @@ -2170,15 +2175,15 @@ static int dmz_to_next_set_block(struct dmz_metadata *zmd, struct dm_zone *zone, /* Get offset */ bitmap = (unsigned long *) mblk->data; bit = chunk_block & DMZ_BLOCK_MASK_BITS; - nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit); + nr_bits = min(nr_blocks, zone_bits - bit); if (set) - set_bit = find_next_bit(bitmap, DMZ_BLOCK_SIZE_BITS, bit); + set_bit = find_next_bit(bitmap, zone_bits, bit); else - set_bit = find_next_zero_bit(bitmap, DMZ_BLOCK_SIZE_BITS, bit); + set_bit = find_next_zero_bit(bitmap, zone_bits, bit); dmz_release_mblock(zmd, mblk); n += set_bit - bit; - if (set_bit < DMZ_BLOCK_SIZE_BITS) + if (set_bit < zone_bits) break; nr_blocks -= nr_bits; @@ -2281,7 +2286,7 @@ static void dmz_get_zone_weight(struct dmz_metadata *zmd, struct dm_zone *zone) /* Count bits in this block */ bitmap = mblk->data; bit = chunk_block & DMZ_BLOCK_MASK_BITS; - nr_bits = min(nr_blocks, DMZ_BLOCK_SIZE_BITS - bit); + nr_bits = min(nr_blocks, zmd->zone_bits_per_mblk - bit); n += dmz_count_bits(bitmap, bit, nr_bits); dmz_release_mblock(zmd, mblk); diff --git a/drivers/md/dm.c b/drivers/md/dm.c index e8f9661a10a1..b89f07ee2eff 100644 --- a/drivers/md/dm.c +++ b/drivers/md/dm.c @@ -1859,6 +1859,7 @@ static void dm_init_normal_md_queue(struct mapped_device *md) /* * Initialize aspects of queue that aren't relevant for blk-mq */ + md->queue->backing_dev_info->congested_data = md; md->queue->backing_dev_info->congested_fn = dm_any_congested; } @@ -1949,7 +1950,12 @@ static struct mapped_device *alloc_dev(int minor) if (!md->queue) goto bad; md->queue->queuedata = md; - md->queue->backing_dev_info->congested_data = md; + /* + * default to bio-based required ->make_request_fn until DM + * table is loaded and md->type established. If request-based + * table is loaded: blk-mq will override accordingly. + */ + blk_queue_make_request(md->queue, dm_make_request); md->disk = alloc_disk_node(1, md->numa_node_id); if (!md->disk) @@ -2264,7 +2270,6 @@ int dm_setup_md_queue(struct mapped_device *md, struct dm_table *t) case DM_TYPE_DAX_BIO_BASED: case DM_TYPE_NVME_BIO_BASED: dm_init_normal_md_queue(md); - blk_queue_make_request(md->queue, dm_make_request); break; case DM_TYPE_NONE: WARN_ON_ONCE(true); diff --git a/drivers/md/md-bitmap.c b/drivers/md/md-bitmap.c index 3ad18246fcb3..b952bd45bd6a 100644 --- a/drivers/md/md-bitmap.c +++ b/drivers/md/md-bitmap.c @@ -364,7 +364,7 @@ static int read_page(struct file *file, unsigned long index, int ret = 0; struct inode *inode = file_inode(file); struct buffer_head *bh; - sector_t block; + sector_t block, blk_cur; pr_debug("read bitmap file (%dB @ %llu)\n", (int)PAGE_SIZE, (unsigned long long)index << PAGE_SHIFT); @@ -375,17 +375,21 @@ static int read_page(struct file *file, unsigned long index, goto out; } attach_page_buffers(page, bh); - block = index << (PAGE_SHIFT - inode->i_blkbits); + blk_cur = index << (PAGE_SHIFT - inode->i_blkbits); while (bh) { + block = blk_cur; + if (count == 0) bh->b_blocknr = 0; else { - bh->b_blocknr = bmap(inode, block); - if (bh->b_blocknr == 0) { - /* Cannot use this file! */ + ret = bmap(inode, &block); + if (ret || !block) { ret = -EINVAL; + bh->b_blocknr = 0; goto out; } + + bh->b_blocknr = block; bh->b_bdev = inode->i_sb->s_bdev; if (count < (1<<inode->i_blkbits)) count = 0; @@ -399,7 +403,7 @@ static int read_page(struct file *file, unsigned long index, set_buffer_mapped(bh); submit_bh(REQ_OP_READ, 0, bh); } - block++; + blk_cur++; bh = bh->b_this_page; } page->index = index; @@ -1019,8 +1023,6 @@ void md_bitmap_unplug(struct bitmap *bitmap) /* look at each page to see if there are any set bits that need to be * flushed out to disk */ for (i = 0; i < bitmap->storage.file_pages; i++) { - if (!bitmap->storage.filemap) - return; dirty = test_and_clear_page_attr(bitmap, i, BITMAP_PAGE_DIRTY); need_write = test_and_clear_page_attr(bitmap, i, BITMAP_PAGE_NEEDWRITE); @@ -1338,7 +1340,8 @@ void md_bitmap_daemon_work(struct mddev *mddev) BITMAP_PAGE_DIRTY)) /* bitmap_unplug will handle the rest */ break; - if (test_and_clear_page_attr(bitmap, j, + if (bitmap->storage.filemap && + test_and_clear_page_attr(bitmap, j, BITMAP_PAGE_NEEDWRITE)) { write_page(bitmap, bitmap->storage.filemap[j], 0); } @@ -1790,8 +1793,8 @@ void md_bitmap_destroy(struct mddev *mddev) return; md_bitmap_wait_behind_writes(mddev); - mempool_destroy(mddev->wb_info_pool); - mddev->wb_info_pool = NULL; + if (!mddev->serialize_policy) + mddev_destroy_serial_pool(mddev, NULL, true); mutex_lock(&mddev->bitmap_info.mutex); spin_lock(&mddev->lock); @@ -1908,7 +1911,7 @@ int md_bitmap_load(struct mddev *mddev) goto out; rdev_for_each(rdev, mddev) - mddev_create_wb_pool(mddev, rdev, true); + mddev_create_serial_pool(mddev, rdev, true); if (mddev_is_clustered(mddev)) md_cluster_ops->load_bitmaps(mddev, mddev->bitmap_info.nodes); @@ -2475,16 +2478,16 @@ backlog_store(struct mddev *mddev, const char *buf, size_t len) if (backlog > COUNTER_MAX) return -EINVAL; mddev->bitmap_info.max_write_behind = backlog; - if (!backlog && mddev->wb_info_pool) { - /* wb_info_pool is not needed if backlog is zero */ - mempool_destroy(mddev->wb_info_pool); - mddev->wb_info_pool = NULL; - } else if (backlog && !mddev->wb_info_pool) { - /* wb_info_pool is needed since backlog is not zero */ + if (!backlog && mddev->serial_info_pool) { + /* serial_info_pool is not needed if backlog is zero */ + if (!mddev->serialize_policy) + mddev_destroy_serial_pool(mddev, NULL, false); + } else if (backlog && !mddev->serial_info_pool) { + /* serial_info_pool is needed since backlog is not zero */ struct md_rdev *rdev; rdev_for_each(rdev, mddev) - mddev_create_wb_pool(mddev, rdev, false); + mddev_create_serial_pool(mddev, rdev, false); } if (old_mwb != backlog) md_bitmap_update_sb(mddev->bitmap); diff --git a/drivers/md/md.c b/drivers/md/md.c index 4e7c9f398bc6..469f551863be 100644 --- a/drivers/md/md.c +++ b/drivers/md/md.c @@ -125,74 +125,165 @@ static inline int speed_max(struct mddev *mddev) mddev->sync_speed_max : sysctl_speed_limit_max; } -static int rdev_init_wb(struct md_rdev *rdev) +static void rdev_uninit_serial(struct md_rdev *rdev) { - if (rdev->bdev->bd_queue->nr_hw_queues == 1) + if (!test_and_clear_bit(CollisionCheck, &rdev->flags)) + return; + + kvfree(rdev->serial); + rdev->serial = NULL; +} + +static void rdevs_uninit_serial(struct mddev *mddev) +{ + struct md_rdev *rdev; + + rdev_for_each(rdev, mddev) + rdev_uninit_serial(rdev); +} + +static int rdev_init_serial(struct md_rdev *rdev) +{ + /* serial_nums equals with BARRIER_BUCKETS_NR */ + int i, serial_nums = 1 << ((PAGE_SHIFT - ilog2(sizeof(atomic_t)))); + struct serial_in_rdev *serial = NULL; + + if (test_bit(CollisionCheck, &rdev->flags)) return 0; - spin_lock_init(&rdev->wb_list_lock); - INIT_LIST_HEAD(&rdev->wb_list); - init_waitqueue_head(&rdev->wb_io_wait); - set_bit(WBCollisionCheck, &rdev->flags); + serial = kvmalloc(sizeof(struct serial_in_rdev) * serial_nums, + GFP_KERNEL); + if (!serial) + return -ENOMEM; - return 1; + for (i = 0; i < serial_nums; i++) { + struct serial_in_rdev *serial_tmp = &serial[i]; + + spin_lock_init(&serial_tmp->serial_lock); + serial_tmp->serial_rb = RB_ROOT_CACHED; + init_waitqueue_head(&serial_tmp->serial_io_wait); + } + + rdev->serial = serial; + set_bit(CollisionCheck, &rdev->flags); + + return 0; +} + +static int rdevs_init_serial(struct mddev *mddev) +{ + struct md_rdev *rdev; + int ret = 0; + + rdev_for_each(rdev, mddev) { + ret = rdev_init_serial(rdev); + if (ret) + break; + } + + /* Free all resources if pool is not existed */ + if (ret && !mddev->serial_info_pool) + rdevs_uninit_serial(mddev); + + return ret; } /* - * Create wb_info_pool if rdev is the first multi-queue device flaged - * with writemostly, also write-behind mode is enabled. + * rdev needs to enable serial stuffs if it meets the conditions: + * 1. it is multi-queue device flaged with writemostly. + * 2. the write-behind mode is enabled. */ -void mddev_create_wb_pool(struct mddev *mddev, struct md_rdev *rdev, - bool is_suspend) +static int rdev_need_serial(struct md_rdev *rdev) { - if (mddev->bitmap_info.max_write_behind == 0) - return; + return (rdev && rdev->mddev->bitmap_info.max_write_behind > 0 && + rdev->bdev->bd_queue->nr_hw_queues != 1 && + test_bit(WriteMostly, &rdev->flags)); +} + +/* + * Init resource for rdev(s), then create serial_info_pool if: + * 1. rdev is the first device which return true from rdev_enable_serial. + * 2. rdev is NULL, means we want to enable serialization for all rdevs. + */ +void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev, + bool is_suspend) +{ + int ret = 0; - if (!test_bit(WriteMostly, &rdev->flags) || !rdev_init_wb(rdev)) + if (rdev && !rdev_need_serial(rdev) && + !test_bit(CollisionCheck, &rdev->flags)) return; - if (mddev->wb_info_pool == NULL) { + if (!is_suspend) + mddev_suspend(mddev); + + if (!rdev) + ret = rdevs_init_serial(mddev); + else + ret = rdev_init_serial(rdev); + if (ret) + goto abort; + + if (mddev->serial_info_pool == NULL) { unsigned int noio_flag; - if (!is_suspend) - mddev_suspend(mddev); noio_flag = memalloc_noio_save(); - mddev->wb_info_pool = mempool_create_kmalloc_pool(NR_WB_INFOS, - sizeof(struct wb_info)); + mddev->serial_info_pool = + mempool_create_kmalloc_pool(NR_SERIAL_INFOS, + sizeof(struct serial_info)); memalloc_noio_restore(noio_flag); - if (!mddev->wb_info_pool) - pr_err("can't alloc memory pool for writemostly\n"); - if (!is_suspend) - mddev_resume(mddev); + if (!mddev->serial_info_pool) { + rdevs_uninit_serial(mddev); + pr_err("can't alloc memory pool for serialization\n"); + } } + +abort: + if (!is_suspend) + mddev_resume(mddev); } -EXPORT_SYMBOL_GPL(mddev_create_wb_pool); /* - * destroy wb_info_pool if rdev is the last device flaged with WBCollisionCheck. + * Free resource from rdev(s), and destroy serial_info_pool under conditions: + * 1. rdev is the last device flaged with CollisionCheck. + * 2. when bitmap is destroyed while policy is not enabled. + * 3. for disable policy, the pool is destroyed only when no rdev needs it. */ -static void mddev_destroy_wb_pool(struct mddev *mddev, struct md_rdev *rdev) +void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev, + bool is_suspend) { - if (!test_and_clear_bit(WBCollisionCheck, &rdev->flags)) + if (rdev && !test_bit(CollisionCheck, &rdev->flags)) return; - if (mddev->wb_info_pool) { + if (mddev->serial_info_pool) { struct md_rdev *temp; - int num = 0; + int num = 0; /* used to track if other rdevs need the pool */ - /* - * Check if other rdevs need wb_info_pool. - */ - rdev_for_each(temp, mddev) - if (temp != rdev && - test_bit(WBCollisionCheck, &temp->flags)) + if (!is_suspend) + mddev_suspend(mddev); + rdev_for_each(temp, mddev) { + if (!rdev) { + if (!mddev->serialize_policy || + !rdev_need_serial(temp)) + rdev_uninit_serial(temp); + else + num++; + } else if (temp != rdev && + test_bit(CollisionCheck, &temp->flags)) num++; - if (!num) { - mddev_suspend(rdev->mddev); - mempool_destroy(mddev->wb_info_pool); - mddev->wb_info_pool = NULL; - mddev_resume(rdev->mddev); } + + if (rdev) + rdev_uninit_serial(rdev); + + if (num) + pr_info("The mempool could be used by other devices\n"); + else { + mempool_destroy(mddev->serial_info_pool); + mddev->serial_info_pool = NULL; + } + if (!is_suspend) + mddev_resume(mddev); } } @@ -2337,7 +2428,7 @@ static int bind_rdev_to_array(struct md_rdev *rdev, struct mddev *mddev) pr_debug("md: bind<%s>\n", b); if (mddev->raid_disks) - mddev_create_wb_pool(mddev, rdev, false); + mddev_create_serial_pool(mddev, rdev, false); if ((err = kobject_add(&rdev->kobj, &mddev->kobj, "dev-%s", b))) goto fail; @@ -2375,7 +2466,7 @@ static void unbind_rdev_from_array(struct md_rdev *rdev) bd_unlink_disk_holder(rdev->bdev, rdev->mddev->gendisk); list_del_rcu(&rdev->same_set); pr_debug("md: unbind<%s>\n", bdevname(rdev->bdev,b)); - mddev_destroy_wb_pool(rdev->mddev, rdev); + mddev_destroy_serial_pool(rdev->mddev, rdev, false); rdev->mddev = NULL; sysfs_remove_link(&rdev->kobj, "block"); sysfs_put(rdev->sysfs_state); @@ -2888,10 +2979,10 @@ state_store(struct md_rdev *rdev, const char *buf, size_t len) } } else if (cmd_match(buf, "writemostly")) { set_bit(WriteMostly, &rdev->flags); - mddev_create_wb_pool(rdev->mddev, rdev, false); + mddev_create_serial_pool(rdev->mddev, rdev, false); err = 0; } else if (cmd_match(buf, "-writemostly")) { - mddev_destroy_wb_pool(rdev->mddev, rdev); + mddev_destroy_serial_pool(rdev->mddev, rdev, false); clear_bit(WriteMostly, &rdev->flags); err = 0; } else if (cmd_match(buf, "blocked")) { @@ -5277,6 +5368,57 @@ static struct md_sysfs_entry md_fail_last_dev = __ATTR(fail_last_dev, S_IRUGO | S_IWUSR, fail_last_dev_show, fail_last_dev_store); +static ssize_t serialize_policy_show(struct mddev *mddev, char *page) +{ + if (mddev->pers == NULL || (mddev->pers->level != 1)) + return sprintf(page, "n/a\n"); + else + return sprintf(page, "%d\n", mddev->serialize_policy); +} + +/* + * Setting serialize_policy to true to enforce write IO is not reordered + * for raid1. + */ +static ssize_t +serialize_policy_store(struct mddev *mddev, const char *buf, size_t len) +{ + int err; + bool value; + + err = kstrtobool(buf, &value); + if (err) + return err; + + if (value == mddev->serialize_policy) + return len; + + err = mddev_lock(mddev); + if (err) + return err; + if (mddev->pers == NULL || (mddev->pers->level != 1)) { + pr_err("md: serialize_policy is only effective for raid1\n"); + err = -EINVAL; + goto unlock; + } + + mddev_suspend(mddev); + if (value) + mddev_create_serial_pool(mddev, NULL, true); + else + mddev_destroy_serial_pool(mddev, NULL, true); + mddev->serialize_policy = value; + mddev_resume(mddev); +unlock: + mddev_unlock(mddev); + return err ?: len; +} + +static struct md_sysfs_entry md_serialize_policy = +__ATTR(serialize_policy, S_IRUGO | S_IWUSR, serialize_policy_show, + serialize_policy_store); + + static struct attribute *md_default_attrs[] = { &md_level.attr, &md_layout.attr, @@ -5294,6 +5436,7 @@ static struct attribute *md_default_attrs[] = { &max_corr_read_errors.attr, &md_consistency_policy.attr, &md_fail_last_dev.attr, + &md_serialize_policy.attr, NULL, }; @@ -5769,18 +5912,18 @@ int md_run(struct mddev *mddev) goto bitmap_abort; if (mddev->bitmap_info.max_write_behind > 0) { - bool creat_pool = false; + bool create_pool = false; rdev_for_each(rdev, mddev) { if (test_bit(WriteMostly, &rdev->flags) && - rdev_init_wb(rdev)) - creat_pool = true; - } - if (creat_pool && mddev->wb_info_pool == NULL) { - mddev->wb_info_pool = - mempool_create_kmalloc_pool(NR_WB_INFOS, - sizeof(struct wb_info)); - if (!mddev->wb_info_pool) { + rdev_init_serial(rdev)) + create_pool = true; + } + if (create_pool && mddev->serial_info_pool == NULL) { + mddev->serial_info_pool = + mempool_create_kmalloc_pool(NR_SERIAL_INFOS, + sizeof(struct serial_info)); + if (!mddev->serial_info_pool) { err = -ENOMEM; goto bitmap_abort; } @@ -6025,8 +6168,9 @@ static void __md_stop_writes(struct mddev *mddev) mddev->in_sync = 1; md_update_sb(mddev, 1); } - mempool_destroy(mddev->wb_info_pool); - mddev->wb_info_pool = NULL; + /* disable policy to guarantee rdevs free resources for serialization */ + mddev->serialize_policy = 0; + mddev_destroy_serial_pool(mddev, NULL, true); } void md_stop_writes(struct mddev *mddev) @@ -8135,13 +8279,12 @@ static __poll_t mdstat_poll(struct file *filp, poll_table *wait) return mask; } -static const struct file_operations md_seq_fops = { - .owner = THIS_MODULE, - .open = md_seq_open, - .read = seq_read, - .llseek = seq_lseek, - .release = seq_release, - .poll = mdstat_poll, +static const struct proc_ops mdstat_proc_ops = { + .proc_open = md_seq_open, + .proc_read = seq_read, + .proc_lseek = seq_lseek, + .proc_release = seq_release, + .proc_poll = mdstat_poll, }; int register_md_personality(struct md_personality *p) @@ -9310,7 +9453,7 @@ static void md_geninit(void) { pr_debug("md: sizeof(mdp_super_t) = %d\n", (int)sizeof(mdp_super_t)); - proc_create("mdstat", S_IRUGO, NULL, &md_seq_fops); + proc_create("mdstat", S_IRUGO, NULL, &mdstat_proc_ops); } static int __init md_init(void) diff --git a/drivers/md/md.h b/drivers/md/md.h index 5f86f8adb0a4..acd681939112 100644 --- a/drivers/md/md.h +++ b/drivers/md/md.h @@ -32,6 +32,16 @@ * be retried. */ #define MD_FAILFAST (REQ_FAILFAST_DEV | REQ_FAILFAST_TRANSPORT) + +/* + * The struct embedded in rdev is used to serialize IO. + */ +struct serial_in_rdev { + struct rb_root_cached serial_rb; + spinlock_t serial_lock; + wait_queue_head_t serial_io_wait; +}; + /* * MD's 'extended' device */ @@ -110,12 +120,7 @@ struct md_rdev { * in superblock. */ - /* - * The members for check collision of write behind IOs. - */ - struct list_head wb_list; - spinlock_t wb_list_lock; - wait_queue_head_t wb_io_wait; + struct serial_in_rdev *serial; /* used for raid1 io serialization */ struct work_struct del_work; /* used for delayed sysfs removal */ @@ -201,9 +206,9 @@ enum flag_bits { * it didn't fail, so don't use FailFast * any more for metadata */ - WBCollisionCheck, /* - * multiqueue device should check if there - * is collision between write behind bios. + CollisionCheck, /* + * check if there is collision between raid1 + * serial bios. */ }; @@ -263,12 +268,13 @@ enum mddev_sb_flags { MD_SB_NEED_REWRITE, /* metadata write needs to be repeated */ }; -#define NR_WB_INFOS 8 -/* record current range of write behind IOs */ -struct wb_info { - sector_t lo; - sector_t hi; - struct list_head list; +#define NR_SERIAL_INFOS 8 +/* record current range of serialize IOs */ +struct serial_info { + struct rb_node node; + sector_t start; /* start sector of rb node */ + sector_t last; /* end sector of rb node */ + sector_t _subtree_last; /* highest sector in subtree of rb node */ }; struct mddev { @@ -487,13 +493,14 @@ struct mddev { */ struct work_struct flush_work; struct work_struct event_work; /* used by dm to report failure event */ - mempool_t *wb_info_pool; + mempool_t *serial_info_pool; void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev); struct md_cluster_info *cluster_info; unsigned int good_device_nr; /* good device num within cluster raid */ bool has_superblocks:1; bool fail_last_dev:1; + bool serialize_policy:1; }; enum recovery_flags { @@ -737,8 +744,10 @@ extern struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs, extern void md_reload_sb(struct mddev *mddev, int raid_disk); extern void md_update_sb(struct mddev *mddev, int force); extern void md_kick_rdev_from_array(struct md_rdev * rdev); -extern void mddev_create_wb_pool(struct mddev *mddev, struct md_rdev *rdev, - bool is_suspend); +extern void mddev_create_serial_pool(struct mddev *mddev, struct md_rdev *rdev, + bool is_suspend); +extern void mddev_destroy_serial_pool(struct mddev *mddev, struct md_rdev *rdev, + bool is_suspend); struct md_rdev *md_find_rdev_nr_rcu(struct mddev *mddev, int nr); struct md_rdev *md_find_rdev_rcu(struct mddev *mddev, dev_t dev); diff --git a/drivers/md/persistent-data/dm-space-map-common.c b/drivers/md/persistent-data/dm-space-map-common.c index bd68f6fef694..d8b4125e338c 100644 --- a/drivers/md/persistent-data/dm-space-map-common.c +++ b/drivers/md/persistent-data/dm-space-map-common.c @@ -380,6 +380,33 @@ int sm_ll_find_free_block(struct ll_disk *ll, dm_block_t begin, return -ENOSPC; } +int sm_ll_find_common_free_block(struct ll_disk *old_ll, struct ll_disk *new_ll, + dm_block_t begin, dm_block_t end, dm_block_t *b) +{ + int r; + uint32_t count; + + do { + r = sm_ll_find_free_block(new_ll, begin, new_ll->nr_blocks, b); + if (r) + break; + + /* double check this block wasn't used in the old transaction */ + if (*b >= old_ll->nr_blocks) + count = 0; + else { + r = sm_ll_lookup(old_ll, *b, &count); + if (r) + break; + + if (count) + begin = *b + 1; + } + } while (count); + + return r; +} + static int sm_ll_mutate(struct ll_disk *ll, dm_block_t b, int (*mutator)(void *context, uint32_t old, uint32_t *new), void *context, enum allocation_event *ev) diff --git a/drivers/md/persistent-data/dm-space-map-common.h b/drivers/md/persistent-data/dm-space-map-common.h index b3078d5eda0c..8de63ce39bdd 100644 --- a/drivers/md/persistent-data/dm-space-map-common.h +++ b/drivers/md/persistent-data/dm-space-map-common.h @@ -109,6 +109,8 @@ int sm_ll_lookup_bitmap(struct ll_disk *ll, dm_block_t b, uint32_t *result); int sm_ll_lookup(struct ll_disk *ll, dm_block_t b, uint32_t *result); int sm_ll_find_free_block(struct ll_disk *ll, dm_block_t begin, dm_block_t end, dm_block_t *result); +int sm_ll_find_common_free_block(struct ll_disk *old_ll, struct ll_disk *new_ll, + dm_block_t begin, dm_block_t end, dm_block_t *result); int sm_ll_insert(struct ll_disk *ll, dm_block_t b, uint32_t ref_count, enum allocation_event *ev); int sm_ll_inc(struct ll_disk *ll, dm_block_t b, enum allocation_event *ev); int sm_ll_dec(struct ll_disk *ll, dm_block_t b, enum allocation_event *ev); diff --git a/drivers/md/persistent-data/dm-space-map-disk.c b/drivers/md/persistent-data/dm-space-map-disk.c index 32adf6b4a9c7..bf4c5e2ccb6f 100644 --- a/drivers/md/persistent-data/dm-space-map-disk.c +++ b/drivers/md/persistent-data/dm-space-map-disk.c @@ -167,8 +167,10 @@ static int sm_disk_new_block(struct dm_space_map *sm, dm_block_t *b) enum allocation_event ev; struct sm_disk *smd = container_of(sm, struct sm_disk, sm); - /* FIXME: we should loop round a couple of times */ - r = sm_ll_find_free_block(&smd->old_ll, smd->begin, smd->old_ll.nr_blocks, b); + /* + * Any block we allocate has to be free in both the old and current ll. + */ + r = sm_ll_find_common_free_block(&smd->old_ll, &smd->ll, smd->begin, smd->ll.nr_blocks, b); if (r) return r; diff --git a/drivers/md/persistent-data/dm-space-map-metadata.c b/drivers/md/persistent-data/dm-space-map-metadata.c index 25328582cc48..9e3c64ec2026 100644 --- a/drivers/md/persistent-data/dm-space-map-metadata.c +++ b/drivers/md/persistent-data/dm-space-map-metadata.c @@ -448,7 +448,10 @@ static int sm_metadata_new_block_(struct dm_space_map *sm, dm_block_t *b) enum allocation_event ev; struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm); - r = sm_ll_find_free_block(&smm->old_ll, smm->begin, smm->old_ll.nr_blocks, b); + /* + * Any block we allocate has to be free in both the old and current ll. + */ + r = sm_ll_find_common_free_block(&smm->old_ll, &smm->ll, smm->begin, smm->ll.nr_blocks, b); if (r) return r; diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c index 201fd8aec59a..cd810e195086 100644 --- a/drivers/md/raid1.c +++ b/drivers/md/raid1.c @@ -29,6 +29,7 @@ #include <linux/module.h> #include <linux/seq_file.h> #include <linux/ratelimit.h> +#include <linux/interval_tree_generic.h> #include <trace/events/block.h> @@ -50,55 +51,71 @@ static void lower_barrier(struct r1conf *conf, sector_t sector_nr); #include "raid1-10.c" -static int check_and_add_wb(struct md_rdev *rdev, sector_t lo, sector_t hi) +#define START(node) ((node)->start) +#define LAST(node) ((node)->last) +INTERVAL_TREE_DEFINE(struct serial_info, node, sector_t, _subtree_last, + START, LAST, static inline, raid1_rb); + +static int check_and_add_serial(struct md_rdev *rdev, struct r1bio *r1_bio, + struct serial_info *si, int idx) { - struct wb_info *wi, *temp_wi; unsigned long flags; int ret = 0; - struct mddev *mddev = rdev->mddev; - - wi = mempool_alloc(mddev->wb_info_pool, GFP_NOIO); - - spin_lock_irqsave(&rdev->wb_list_lock, flags); - list_for_each_entry(temp_wi, &rdev->wb_list, list) { - /* collision happened */ - if (hi > temp_wi->lo && lo < temp_wi->hi) { - ret = -EBUSY; - break; - } + sector_t lo = r1_bio->sector; + sector_t hi = lo + r1_bio->sectors; + struct serial_in_rdev *serial = &rdev->serial[idx]; + + spin_lock_irqsave(&serial->serial_lock, flags); + /* collision happened */ + if (raid1_rb_iter_first(&serial->serial_rb, lo, hi)) + ret = -EBUSY; + else { + si->start = lo; + si->last = hi; + raid1_rb_insert(si, &serial->serial_rb); } - - if (!ret) { - wi->lo = lo; - wi->hi = hi; - list_add(&wi->list, &rdev->wb_list); - } else - mempool_free(wi, mddev->wb_info_pool); - spin_unlock_irqrestore(&rdev->wb_list_lock, flags); + spin_unlock_irqrestore(&serial->serial_lock, flags); return ret; } -static void remove_wb(struct md_rdev *rdev, sector_t lo, sector_t hi) +static void wait_for_serialization(struct md_rdev *rdev, struct r1bio *r1_bio) +{ + struct mddev *mddev = rdev->mddev; + struct serial_info *si; + int idx = sector_to_idx(r1_bio->sector); + struct serial_in_rdev *serial = &rdev->serial[idx]; + + if (WARN_ON(!mddev->serial_info_pool)) + return; + si = mempool_alloc(mddev->serial_info_pool, GFP_NOIO); + wait_event(serial->serial_io_wait, + check_and_add_serial(rdev, r1_bio, si, idx) == 0); +} + +static void remove_serial(struct md_rdev *rdev, sector_t lo, sector_t hi) { - struct wb_info *wi; + struct serial_info *si; unsigned long flags; int found = 0; struct mddev *mddev = rdev->mddev; - - spin_lock_irqsave(&rdev->wb_list_lock, flags); - list_for_each_entry(wi, &rdev->wb_list, list) - if (hi == wi->hi && lo == wi->lo) { - list_del(&wi->list); - mempool_free(wi, mddev->wb_info_pool); + int idx = sector_to_idx(lo); + struct serial_in_rdev *serial = &rdev->serial[idx]; + + spin_lock_irqsave(&serial->serial_lock, flags); + for (si = raid1_rb_iter_first(&serial->serial_rb, lo, hi); + si; si = raid1_rb_iter_next(si, lo, hi)) { + if (si->start == lo && si->last == hi) { + raid1_rb_remove(si, &serial->serial_rb); + mempool_free(si, mddev->serial_info_pool); found = 1; break; } - + } if (!found) - WARN(1, "The write behind IO is not recorded\n"); - spin_unlock_irqrestore(&rdev->wb_list_lock, flags); - wake_up(&rdev->wb_io_wait); + WARN(1, "The write IO is not recorded for serialization\n"); + spin_unlock_irqrestore(&serial->serial_lock, flags); + wake_up(&serial->serial_io_wait); } /* @@ -430,6 +447,8 @@ static void raid1_end_write_request(struct bio *bio) int mirror = find_bio_disk(r1_bio, bio); struct md_rdev *rdev = conf->mirrors[mirror].rdev; bool discard_error; + sector_t lo = r1_bio->sector; + sector_t hi = r1_bio->sector + r1_bio->sectors; discard_error = bio->bi_status && bio_op(bio) == REQ_OP_DISCARD; @@ -499,12 +518,8 @@ static void raid1_end_write_request(struct bio *bio) } if (behind) { - if (test_bit(WBCollisionCheck, &rdev->flags)) { - sector_t lo = r1_bio->sector; - sector_t hi = r1_bio->sector + r1_bio->sectors; - - remove_wb(rdev, lo, hi); - } + if (test_bit(CollisionCheck, &rdev->flags)) + remove_serial(rdev, lo, hi); if (test_bit(WriteMostly, &rdev->flags)) atomic_dec(&r1_bio->behind_remaining); @@ -527,7 +542,8 @@ static void raid1_end_write_request(struct bio *bio) call_bio_endio(r1_bio); } } - } + } else if (rdev->mddev->serialize_policy) + remove_serial(rdev, lo, hi); if (r1_bio->bios[mirror] == NULL) rdev_dec_pending(rdev, conf->mddev); @@ -1479,6 +1495,7 @@ static void raid1_write_request(struct mddev *mddev, struct bio *bio, for (i = 0; i < disks; i++) { struct bio *mbio = NULL; + struct md_rdev *rdev = conf->mirrors[i].rdev; if (!r1_bio->bios[i]) continue; @@ -1506,18 +1523,12 @@ static void raid1_write_request(struct mddev *mddev, struct bio *bio, mbio = bio_clone_fast(bio, GFP_NOIO, &mddev->bio_set); if (r1_bio->behind_master_bio) { - struct md_rdev *rdev = conf->mirrors[i].rdev; - - if (test_bit(WBCollisionCheck, &rdev->flags)) { - sector_t lo = r1_bio->sector; - sector_t hi = r1_bio->sector + r1_bio->sectors; - - wait_event(rdev->wb_io_wait, - check_and_add_wb(rdev, lo, hi) == 0); - } + if (test_bit(CollisionCheck, &rdev->flags)) + wait_for_serialization(rdev, r1_bio); if (test_bit(WriteMostly, &rdev->flags)) atomic_inc(&r1_bio->behind_remaining); - } + } else if (mddev->serialize_policy) + wait_for_serialization(rdev, r1_bio); r1_bio->bios[i] = mbio; diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c index d4d3b67ffbba..ba00e9877f02 100644 --- a/drivers/md/raid5.c +++ b/drivers/md/raid5.c @@ -6598,7 +6598,6 @@ raid5_show_group_thread_cnt(struct mddev *mddev, char *page) static int alloc_thread_groups(struct r5conf *conf, int cnt, int *group_cnt, - int *worker_cnt_per_group, struct r5worker_group **worker_groups); static ssize_t raid5_store_group_thread_cnt(struct mddev *mddev, const char *page, size_t len) @@ -6607,7 +6606,7 @@ raid5_store_group_thread_cnt(struct mddev *mddev, const char *page, size_t len) unsigned int new; int err; struct r5worker_group *new_groups, *old_groups; - int group_cnt, worker_cnt_per_group; + int group_cnt; if (len >= PAGE_SIZE) return -EINVAL; @@ -6630,13 +6629,11 @@ raid5_store_group_thread_cnt(struct mddev *mddev, const char *page, size_t len) if (old_groups) flush_workqueue(raid5_wq); - err = alloc_thread_groups(conf, new, - &group_cnt, &worker_cnt_per_group, - &new_groups); + err = alloc_thread_groups(conf, new, &group_cnt, &new_groups); if (!err) { spin_lock_irq(&conf->device_lock); conf->group_cnt = group_cnt; - conf->worker_cnt_per_group = worker_cnt_per_group; + conf->worker_cnt_per_group = new; conf->worker_groups = new_groups; spin_unlock_irq(&conf->device_lock); @@ -6672,16 +6669,13 @@ static struct attribute_group raid5_attrs_group = { .attrs = raid5_attrs, }; -static int alloc_thread_groups(struct r5conf *conf, int cnt, - int *group_cnt, - int *worker_cnt_per_group, +static int alloc_thread_groups(struct r5conf *conf, int cnt, int *group_cnt, struct r5worker_group **worker_groups) { int i, j, k; ssize_t size; struct r5worker *workers; - *worker_cnt_per_group = cnt; if (cnt == 0) { *group_cnt = 0; *worker_groups = NULL; @@ -6882,7 +6876,7 @@ static struct r5conf *setup_conf(struct mddev *mddev) struct disk_info *disk; char pers_name[6]; int i; - int group_cnt, worker_cnt_per_group; + int group_cnt; struct r5worker_group *new_group; int ret; @@ -6928,10 +6922,9 @@ static struct r5conf *setup_conf(struct mddev *mddev) for (i = 0; i < PENDING_IO_MAX; i++) list_add(&conf->pending_data[i].sibling, &conf->free_list); /* Don't enable multi-threading by default*/ - if (!alloc_thread_groups(conf, 0, &group_cnt, &worker_cnt_per_group, - &new_group)) { + if (!alloc_thread_groups(conf, 0, &group_cnt, &new_group)) { conf->group_cnt = group_cnt; - conf->worker_cnt_per_group = worker_cnt_per_group; + conf->worker_cnt_per_group = 0; conf->worker_groups = new_group; } else goto abort; |