bcache: remove embedded struct cache_sb from struct cache_set
authorColy Li <colyli@suse.de>
Thu, 1 Oct 2020 06:50:56 +0000 (14:50 +0800)
committerJens Axboe <axboe@kernel.dk>
Fri, 2 Oct 2020 20:25:30 +0000 (14:25 -0600)
Since bcache code was merged into mainline kerrnel, each cache set only
as one single cache in it. The multiple caches framework is here but the
code is far from completed. Considering the multiple copies of cached
data can also be stored on e.g. md raid1 devices, it is unnecessary to
support multiple caches in one cache set indeed.

The previous preparation patches fix the dependencies of explicitly
making a cache set only have single cache. Now we don't have to maintain
an embedded partial super block in struct cache_set, the in-memory super
block can be directly referenced from struct cache.

This patch removes the embedded struct cache_sb from struct cache_set,
and fixes all locations where the superb lock was referenced from this
removed super block by referencing the in-memory super block of struct
cache.

Signed-off-by: Coly Li <colyli@suse.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/md/bcache/alloc.c
drivers/md/bcache/bcache.h
drivers/md/bcache/btree.c
drivers/md/bcache/btree.h
drivers/md/bcache/extents.c
drivers/md/bcache/features.c
drivers/md/bcache/io.c
drivers/md/bcache/journal.c
drivers/md/bcache/request.c
drivers/md/bcache/super.c
drivers/md/bcache/writeback.c

index 65fdbde..8c371d5 100644 (file)
@@ -87,7 +87,7 @@ void bch_rescale_priorities(struct cache_set *c, int sectors)
 {
        struct cache *ca;
        struct bucket *b;
-       unsigned long next = c->nbuckets * c->sb.bucket_size / 1024;
+       unsigned long next = c->nbuckets * c->cache->sb.bucket_size / 1024;
        int r;
 
        atomic_sub(sectors, &c->rescale);
@@ -583,7 +583,7 @@ static struct open_bucket *pick_data_bucket(struct cache_set *c,
                                           struct open_bucket, list);
 found:
        if (!ret->sectors_free && KEY_PTRS(alloc)) {
-               ret->sectors_free = c->sb.bucket_size;
+               ret->sectors_free = c->cache->sb.bucket_size;
                bkey_copy(&ret->key, alloc);
                bkey_init(alloc);
        }
@@ -677,7 +677,7 @@ bool bch_alloc_sectors(struct cache_set *c,
                                &PTR_CACHE(c, &b->key, i)->sectors_written);
        }
 
-       if (b->sectors_free < c->sb.block_size)
+       if (b->sectors_free < c->cache->sb.block_size)
                b->sectors_free = 0;
 
        /*
index 94d4baf..1d57f48 100644 (file)
@@ -517,8 +517,6 @@ struct cache_set {
        atomic_t                idle_counter;
        atomic_t                at_max_writeback_rate;
 
-       struct cache_sb         sb;
-
        struct cache            *cache;
 
        struct bcache_device    **devices;
@@ -799,7 +797,7 @@ static inline sector_t bucket_to_sector(struct cache_set *c, size_t b)
 
 static inline sector_t bucket_remainder(struct cache_set *c, sector_t s)
 {
-       return s & (c->sb.bucket_size - 1);
+       return s & (c->cache->sb.bucket_size - 1);
 }
 
 static inline struct cache *PTR_CACHE(struct cache_set *c,
index 1f3d487..910df24 100644 (file)
@@ -117,7 +117,7 @@ static void bch_btree_init_next(struct btree *b)
 
        if (b->written < btree_blocks(b))
                bch_bset_init_next(&b->keys, write_block(b),
-                                  bset_magic(&b->c->sb));
+                                  bset_magic(&b->c->cache->sb));
 
 }
 
@@ -155,7 +155,7 @@ void bch_btree_node_read_done(struct btree *b)
         * See the comment arount cache_set->fill_iter.
         */
        iter = mempool_alloc(&b->c->fill_iter, GFP_NOIO);
-       iter->size = b->c->sb.bucket_size / b->c->sb.block_size;
+       iter->size = b->c->cache->sb.bucket_size / b->c->cache->sb.block_size;
        iter->used = 0;
 
 #ifdef CONFIG_BCACHE_DEBUG
@@ -178,7 +178,7 @@ void bch_btree_node_read_done(struct btree *b)
                        goto err;
 
                err = "bad magic";
-               if (i->magic != bset_magic(&b->c->sb))
+               if (i->magic != bset_magic(&b->c->cache->sb))
                        goto err;
 
                err = "bad checksum";
@@ -219,7 +219,7 @@ void bch_btree_node_read_done(struct btree *b)
 
        if (b->written < btree_blocks(b))
                bch_bset_init_next(&b->keys, write_block(b),
-                                  bset_magic(&b->c->sb));
+                                  bset_magic(&b->c->cache->sb));
 out:
        mempool_free(iter, &b->c->fill_iter);
        return;
@@ -423,7 +423,7 @@ void __bch_btree_node_write(struct btree *b, struct closure *parent)
 
        do_btree_node_write(b);
 
-       atomic_long_add(set_blocks(i, block_bytes(b->c->cache)) * b->c->sb.block_size,
+       atomic_long_add(set_blocks(i, block_bytes(b->c->cache)) * b->c->cache->sb.block_size,
                        &PTR_CACHE(b->c, &b->key, 0)->btree_sectors_written);
 
        b->written += set_blocks(i, block_bytes(b->c->cache));
@@ -738,7 +738,7 @@ void bch_btree_cache_free(struct cache_set *c)
        if (c->verify_data)
                list_move(&c->verify_data->list, &c->btree_cache);
 
-       free_pages((unsigned long) c->verify_ondisk, ilog2(meta_bucket_pages(&c->sb)));
+       free_pages((unsigned long) c->verify_ondisk, ilog2(meta_bucket_pages(&c->cache->sb)));
 #endif
 
        list_splice(&c->btree_cache_freeable,
@@ -785,7 +785,8 @@ int bch_btree_cache_alloc(struct cache_set *c)
        mutex_init(&c->verify_lock);
 
        c->verify_ondisk = (void *)
-               __get_free_pages(GFP_KERNEL|__GFP_COMP, ilog2(meta_bucket_pages(&c->sb)));
+               __get_free_pages(GFP_KERNEL|__GFP_COMP,
+                                ilog2(meta_bucket_pages(&c->cache->sb)));
        if (!c->verify_ondisk) {
                /*
                 * Don't worry about the mca_rereserve buckets
@@ -1108,7 +1109,7 @@ retry:
        }
 
        b->parent = parent;
-       bch_bset_init_next(&b->keys, b->keys.set->data, bset_magic(&b->c->sb));
+       bch_bset_init_next(&b->keys, b->keys.set->data, bset_magic(&b->c->cache->sb));
 
        mutex_unlock(&c->bucket_lock);
 
index 2579699..5048210 100644 (file)
@@ -194,7 +194,7 @@ static inline unsigned int bset_block_offset(struct btree *b, struct bset *i)
 
 static inline void set_gc_sectors(struct cache_set *c)
 {
-       atomic_set(&c->sectors_to_gc, c->sb.bucket_size * c->nbuckets / 16);
+       atomic_set(&c->sectors_to_gc, c->cache->sb.bucket_size * c->nbuckets / 16);
 }
 
 void bkey_put(struct cache_set *c, struct bkey *k);
index 9162af5..f4658a1 100644 (file)
@@ -54,7 +54,7 @@ static bool __ptr_invalid(struct cache_set *c, const struct bkey *k)
                        size_t bucket = PTR_BUCKET_NR(c, k, i);
                        size_t r = bucket_remainder(c, PTR_OFFSET(k, i));
 
-                       if (KEY_SIZE(k) + r > c->sb.bucket_size ||
+                       if (KEY_SIZE(k) + r > c->cache->sb.bucket_size ||
                            bucket <  ca->sb.first_bucket ||
                            bucket >= ca->sb.nbuckets)
                                return true;
@@ -75,7 +75,7 @@ static const char *bch_ptr_status(struct cache_set *c, const struct bkey *k)
                        size_t bucket = PTR_BUCKET_NR(c, k, i);
                        size_t r = bucket_remainder(c, PTR_OFFSET(k, i));
 
-                       if (KEY_SIZE(k) + r > c->sb.bucket_size)
+                       if (KEY_SIZE(k) + r > c->cache->sb.bucket_size)
                                return "bad, length too big";
                        if (bucket <  ca->sb.first_bucket)
                                return "bad, short offset";
@@ -136,7 +136,7 @@ static void bch_bkey_dump(struct btree_keys *keys, const struct bkey *k)
                size_t n = PTR_BUCKET_NR(b->c, k, j);
 
                pr_cont(" bucket %zu", n);
-               if (n >= b->c->sb.first_bucket && n < b->c->sb.nbuckets)
+               if (n >= b->c->cache->sb.first_bucket && n < b->c->cache->sb.nbuckets)
                        pr_cont(" prio %i",
                                PTR_BUCKET(b->c, k, j)->prio);
        }
index 4442df4..6469223 100644 (file)
@@ -30,7 +30,7 @@ static struct feature feature_list[] = {
        for (f = &feature_list[0]; f->compat != 0; f++) {               \
                if (f->compat != BCH_FEATURE_ ## type)                  \
                        continue;                                       \
-               if (BCH_HAS_ ## type ## _FEATURE(&c->sb, f->mask)) {    \
+               if (BCH_HAS_ ## type ## _FEATURE(&c->cache->sb, f->mask)) {     \
                        if (first) {                                    \
                                out += snprintf(out, buf + size - out,  \
                                                "[");   \
@@ -44,7 +44,7 @@ static struct feature feature_list[] = {
                                                                        \
                out += snprintf(out, buf + size - out, "%s", f->string);\
                                                                        \
-               if (BCH_HAS_ ## type ## _FEATURE(&c->sb, f->mask))      \
+               if (BCH_HAS_ ## type ## _FEATURE(&c->cache->sb, f->mask))       \
                        out += snprintf(out, buf + size - out, "]");    \
                                                                        \
                first = false;                                          \
index a14a445..dad71a6 100644 (file)
@@ -26,7 +26,7 @@ struct bio *bch_bbio_alloc(struct cache_set *c)
        struct bbio *b = mempool_alloc(&c->bio_meta, GFP_NOIO);
        struct bio *bio = &b->bio;
 
-       bio_init(bio, bio->bi_inline_vecs, meta_bucket_pages(&c->sb));
+       bio_init(bio, bio->bi_inline_vecs, meta_bucket_pages(&c->cache->sb));
 
        return bio;
 }
index cd46f33..aefbdb7 100644 (file)
@@ -666,7 +666,7 @@ static void journal_reclaim(struct cache_set *c)
 
        bkey_init(k);
        SET_KEY_PTRS(k, 1);
-       c->journal.blocks_free = c->sb.bucket_size >> c->block_bits;
+       c->journal.blocks_free = ca->sb.bucket_size >> c->block_bits;
 
 out:
        if (!journal_full(&c->journal))
@@ -735,7 +735,7 @@ static void journal_write_unlocked(struct closure *cl)
        struct journal_write *w = c->journal.cur;
        struct bkey *k = &c->journal.key;
        unsigned int i, sectors = set_blocks(w->data, block_bytes(ca)) *
-               c->sb.block_size;
+               ca->sb.block_size;
 
        struct bio *bio;
        struct bio_list list;
@@ -762,7 +762,7 @@ static void journal_write_unlocked(struct closure *cl)
        bkey_copy(&w->data->uuid_bucket, &c->uuid_bucket);
 
        w->data->prio_bucket[ca->sb.nr_this_dev] = ca->prio_buckets[0];
-       w->data->magic          = jset_magic(&c->sb);
+       w->data->magic          = jset_magic(&ca->sb);
        w->data->version        = BCACHE_JSET_VERSION;
        w->data->last_seq       = last_seq(&c->journal);
        w->data->csum           = csum_set(w->data);
@@ -838,6 +838,7 @@ static struct journal_write *journal_wait_for_write(struct cache_set *c,
        size_t sectors;
        struct closure cl;
        bool wait = false;
+       struct cache *ca = c->cache;
 
        closure_init_stack(&cl);
 
@@ -847,10 +848,10 @@ static struct journal_write *journal_wait_for_write(struct cache_set *c,
                struct journal_write *w = c->journal.cur;
 
                sectors = __set_blocks(w->data, w->data->keys + nkeys,
-                                      block_bytes(c->cache)) * c->sb.block_size;
+                                      block_bytes(ca)) * ca->sb.block_size;
 
                if (sectors <= min_t(size_t,
-                                    c->journal.blocks_free * c->sb.block_size,
+                                    c->journal.blocks_free * ca->sb.block_size,
                                     PAGE_SECTORS << JSET_BITS))
                        return w;
 
index d90319d..2143263 100644 (file)
@@ -394,8 +394,8 @@ static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
                        goto skip;
        }
 
-       if (bio->bi_iter.bi_sector & (c->sb.block_size - 1) ||
-           bio_sectors(bio) & (c->sb.block_size - 1)) {
+       if (bio->bi_iter.bi_sector & (c->cache->sb.block_size - 1) ||
+           bio_sectors(bio) & (c->cache->sb.block_size - 1)) {
                pr_debug("skipping unaligned io\n");
                goto skip;
        }
index 9632340..46a0013 100644 (file)
@@ -350,16 +350,10 @@ void bcache_write_super(struct cache_set *c)
        down(&c->sb_write_mutex);
        closure_init(cl, &c->cl);
 
-       c->sb.seq++;
+       ca->sb.seq++;
 
-       if (c->sb.version > version)
-               version = c->sb.version;
-
-       ca->sb.version          = version;
-       ca->sb.seq              = c->sb.seq;
-       ca->sb.last_mount       = c->sb.last_mount;
-
-       SET_CACHE_SYNC(&ca->sb, CACHE_SYNC(&c->sb));
+       if (ca->sb.version < version)
+               ca->sb.version = version;
 
        bio_init(bio, ca->sb_bv, 1);
        bio_set_dev(bio, ca->bdev);
@@ -477,7 +471,7 @@ static int __uuid_write(struct cache_set *c)
 {
        BKEY_PADDED(key) k;
        struct closure cl;
-       struct cache *ca;
+       struct cache *ca = c->cache;
        unsigned int size;
 
        closure_init_stack(&cl);
@@ -486,13 +480,12 @@ static int __uuid_write(struct cache_set *c)
        if (bch_bucket_alloc_set(c, RESERVE_BTREE, &k.key, true))
                return 1;
 
-       size =  meta_bucket_pages(&c->sb) * PAGE_SECTORS;
+       size =  meta_bucket_pages(&ca->sb) * PAGE_SECTORS;
        SET_KEY_SIZE(&k.key, size);
        uuid_io(c, REQ_OP_WRITE, 0, &k.key, &cl);
        closure_sync(&cl);
 
        /* Only one bucket used for uuid write */
-       ca = PTR_CACHE(c, &k.key, 0);
        atomic_long_add(ca->sb.bucket_size, &ca->meta_sectors_written);
 
        bkey_copy(&c->uuid_bucket, &k.key);
@@ -1205,7 +1198,7 @@ int bch_cached_dev_attach(struct cached_dev *dc, struct cache_set *c,
                return -EINVAL;
        }
 
-       if (dc->sb.block_size < c->sb.block_size) {
+       if (dc->sb.block_size < c->cache->sb.block_size) {
                /* Will die */
                pr_err("Couldn't attach %s: block size less than set's block size\n",
                       dc->backing_dev_name);
@@ -1663,6 +1656,9 @@ static void cache_set_free(struct closure *cl)
        bch_journal_free(c);
 
        mutex_lock(&bch_register_lock);
+       bch_bset_sort_state_free(&c->sort);
+       free_pages((unsigned long) c->uuids, ilog2(meta_bucket_pages(&c->cache->sb)));
+
        ca = c->cache;
        if (ca) {
                ca->set = NULL;
@@ -1670,8 +1666,6 @@ static void cache_set_free(struct closure *cl)
                kobject_put(&ca->kobj);
        }
 
-       bch_bset_sort_state_free(&c->sort);
-       free_pages((unsigned long) c->uuids, ilog2(meta_bucket_pages(&c->sb)));
 
        if (c->moving_gc_wq)
                destroy_workqueue(c->moving_gc_wq);
@@ -1837,6 +1831,7 @@ void bch_cache_set_unregister(struct cache_set *c)
 struct cache_set *bch_cache_set_alloc(struct cache_sb *sb)
 {
        int iter_size;
+       struct cache *ca = container_of(sb, struct cache, sb);
        struct cache_set *c = kzalloc(sizeof(struct cache_set), GFP_KERNEL);
 
        if (!c)
@@ -1859,23 +1854,15 @@ struct cache_set *bch_cache_set_alloc(struct cache_sb *sb)
        bch_cache_accounting_init(&c->accounting, &c->cl);
 
        memcpy(c->set_uuid, sb->set_uuid, 16);
-       c->sb.block_size        = sb->block_size;
-       c->sb.bucket_size       = sb->bucket_size;
-       c->sb.nr_in_set         = sb->nr_in_set;
-       c->sb.last_mount        = sb->last_mount;
-       c->sb.version           = sb->version;
-       if (c->sb.version >= BCACHE_SB_VERSION_CDEV_WITH_FEATURES) {
-               c->sb.feature_compat = sb->feature_compat;
-               c->sb.feature_ro_compat = sb->feature_ro_compat;
-               c->sb.feature_incompat = sb->feature_incompat;
-       }
 
+       c->cache                = ca;
+       c->cache->set           = c;
        c->bucket_bits          = ilog2(sb->bucket_size);
        c->block_bits           = ilog2(sb->block_size);
-       c->nr_uuids             = meta_bucket_bytes(&c->sb) / sizeof(struct uuid_entry);
+       c->nr_uuids             = meta_bucket_bytes(sb) / sizeof(struct uuid_entry);
        c->devices_max_used     = 0;
        atomic_set(&c->attached_dev_nr, 0);
-       c->btree_pages          = meta_bucket_pages(&c->sb);
+       c->btree_pages          = meta_bucket_pages(sb);
        if (c->btree_pages > BTREE_MAX_PAGES)
                c->btree_pages = max_t(int, c->btree_pages / 4,
                                       BTREE_MAX_PAGES);
@@ -1913,7 +1900,7 @@ struct cache_set *bch_cache_set_alloc(struct cache_sb *sb)
 
        if (mempool_init_kmalloc_pool(&c->bio_meta, 2,
                        sizeof(struct bbio) +
-                       sizeof(struct bio_vec) * meta_bucket_pages(&c->sb)))
+                       sizeof(struct bio_vec) * meta_bucket_pages(sb)))
                goto err;
 
        if (mempool_init_kmalloc_pool(&c->fill_iter, 1, iter_size))
@@ -1923,7 +1910,7 @@ struct cache_set *bch_cache_set_alloc(struct cache_sb *sb)
                        BIOSET_NEED_BVECS|BIOSET_NEED_RESCUER))
                goto err;
 
-       c->uuids = alloc_meta_bucket_pages(GFP_KERNEL, &c->sb);
+       c->uuids = alloc_meta_bucket_pages(GFP_KERNEL, sb);
        if (!c->uuids)
                goto err;
 
@@ -2103,7 +2090,7 @@ static int run_cache_set(struct cache_set *c)
                goto err;
 
        closure_sync(&cl);
-       c->sb.last_mount = (u32)ktime_get_real_seconds();
+       c->cache->sb.last_mount = (u32)ktime_get_real_seconds();
        bcache_write_super(c);
 
        list_for_each_entry_safe(dc, t, &uncached_devices, list)
index 4f4ad6b..3c74996 100644 (file)
@@ -35,7 +35,7 @@ static uint64_t __calc_target_rate(struct cached_dev *dc)
         * This is the size of the cache, minus the amount used for
         * flash-only devices
         */
-       uint64_t cache_sectors = c->nbuckets * c->sb.bucket_size -
+       uint64_t cache_sectors = c->nbuckets * c->cache->sb.bucket_size -
                                atomic_long_read(&c->flash_dev_dirty_sectors);
 
        /*