1 files changed, 242 insertions, 46 deletions

diff --git a/fs/btrfs/zoned.c b/fs/btrfs/zoned.c
index 69d03feea4e0..9430b34d3cbb 100644
--- a/fs/btrfs/zoned.c
+++ b/fs/btrfs/zoned.c
@@ -707,11 +707,14 @@ int btrfs_check_zoned_mode(struct btrfs_fs_info *fs_info)
 		 * zoned mode. In this case, we don't have a valid max zone
 		 * append size.
 		 */
-		if (bdev_is_zoned(device->bdev)) {
-			blk_stack_limits(lim,
-					 &bdev_get_queue(device->bdev)->limits,
-					 0);
-		}
+		if (bdev_is_zoned(device->bdev))
+			blk_stack_limits(lim, bdev_limits(device->bdev), 0);
+	}
+
+	ret = blk_validate_limits(lim);
+	if (ret) {
+		btrfs_err(fs_info, "zoned: failed to validate queue limits");
+		return ret;
 	}
 
 	/*
@@ -745,8 +748,9 @@ int btrfs_check_zoned_mode(struct btrfs_fs_info *fs_info)
 		     (u64)lim->max_segments << PAGE_SHIFT),
 		fs_info->sectorsize);
 	fs_info->fs_devices->chunk_alloc_policy = BTRFS_CHUNK_ALLOC_ZONED;
-	if (fs_info->max_zone_append_size < fs_info->max_extent_size)
-		fs_info->max_extent_size = fs_info->max_zone_append_size;
+
+	fs_info->max_extent_size = min_not_zero(fs_info->max_extent_size,
+						fs_info->max_zone_append_size);
 
 	/*
 	 * Check mount options here, because we might change fs_info->zoned
@@ -985,7 +989,7 @@ int btrfs_advance_sb_log(struct btrfs_device *device, int mirror)
 	}
 
 	/* All the zones are FULL. Should not reach here. */
-	ASSERT(0);
+	DEBUG_WARN("unexpected state, all zones full");
 	return -EIO;
 }
 
@@ -1273,7 +1277,7 @@ struct zone_info {
 
 static int btrfs_load_zone_info(struct btrfs_fs_info *fs_info, int zone_idx,
 				struct zone_info *info, unsigned long *active,
-				struct btrfs_chunk_map *map)
+				struct btrfs_chunk_map *map, bool new)
 {
 	struct btrfs_dev_replace *dev_replace = &fs_info->dev_replace;
 	struct btrfs_device *device;
@@ -1303,6 +1307,8 @@ static int btrfs_load_zone_info(struct btrfs_fs_info *fs_info, int zone_idx,
 		return 0;
 	}
 
+	ASSERT(!new || btrfs_dev_is_empty_zone(device, info->physical));
+
 	/* This zone will be used for allocation, so mark this zone non-empty. */
 	btrfs_dev_clear_zone_empty(device, info->physical);
 
@@ -1315,6 +1321,18 @@ static int btrfs_load_zone_info(struct btrfs_fs_info *fs_info, int zone_idx,
 	 * to determine the allocation offset within the zone.
 	 */
 	WARN_ON(!IS_ALIGNED(info->physical, fs_info->zone_size));
+
+	if (new) {
+		sector_t capacity;
+
+		capacity = bdev_zone_capacity(device->bdev, info->physical >> SECTOR_SHIFT);
+		up_read(&dev_replace->rwsem);
+		info->alloc_offset = 0;
+		info->capacity = capacity << SECTOR_SHIFT;
+
+		return 0;
+	}
+
 	nofs_flag = memalloc_nofs_save();
 	ret = btrfs_get_dev_zone(device, info->physical, &zone);
 	memalloc_nofs_restore(nofs_flag);
@@ -1385,7 +1403,8 @@ static int btrfs_load_block_group_single(struct btrfs_block_group *bg,
 static int btrfs_load_block_group_dup(struct btrfs_block_group *bg,
 				      struct btrfs_chunk_map *map,
 				      struct zone_info *zone_info,
-				      unsigned long *active)
+				      unsigned long *active,
+				      u64 last_alloc)
 {
 	struct btrfs_fs_info *fs_info = bg->fs_info;
 
@@ -1408,6 +1427,13 @@ static int btrfs_load_block_group_dup(struct btrfs_block_group *bg,
 			  zone_info[1].physical);
 		return -EIO;
 	}
+
+	if (zone_info[0].alloc_offset == WP_CONVENTIONAL)
+		zone_info[0].alloc_offset = last_alloc;
+
+	if (zone_info[1].alloc_offset == WP_CONVENTIONAL)
+		zone_info[1].alloc_offset = last_alloc;
+
 	if (zone_info[0].alloc_offset != zone_info[1].alloc_offset) {
 		btrfs_err(bg->fs_info,
 			  "zoned: write pointer offset mismatch of zones in DUP profile");
@@ -1428,7 +1454,8 @@ static int btrfs_load_block_group_dup(struct btrfs_block_group *bg,
 static int btrfs_load_block_group_raid1(struct btrfs_block_group *bg,
 					struct btrfs_chunk_map *map,
 					struct zone_info *zone_info,
-					unsigned long *active)
+					unsigned long *active,
+					u64 last_alloc)
 {
 	struct btrfs_fs_info *fs_info = bg->fs_info;
 	int i;
@@ -1443,10 +1470,12 @@ static int btrfs_load_block_group_raid1(struct btrfs_block_group *bg,
 	bg->zone_capacity = min_not_zero(zone_info[0].capacity, zone_info[1].capacity);
 
 	for (i = 0; i < map->num_stripes; i++) {
-		if (zone_info[i].alloc_offset == WP_MISSING_DEV ||
-		    zone_info[i].alloc_offset == WP_CONVENTIONAL)
+		if (zone_info[i].alloc_offset == WP_MISSING_DEV)
 			continue;
 
+		if (zone_info[i].alloc_offset == WP_CONVENTIONAL)
+			zone_info[i].alloc_offset = last_alloc;
+
 		if ((zone_info[0].alloc_offset != zone_info[i].alloc_offset) &&
 		    !btrfs_test_opt(fs_info, DEGRADED)) {
 			btrfs_err(fs_info,
@@ -1476,7 +1505,8 @@ static int btrfs_load_block_group_raid1(struct btrfs_block_group *bg,
 static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg,
 					struct btrfs_chunk_map *map,
 					struct zone_info *zone_info,
-					unsigned long *active)
+					unsigned long *active,
+					u64 last_alloc)
 {
 	struct btrfs_fs_info *fs_info = bg->fs_info;
 
@@ -1487,10 +1517,29 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg,
 	}
 
 	for (int i = 0; i < map->num_stripes; i++) {
-		if (zone_info[i].alloc_offset == WP_MISSING_DEV ||
-		    zone_info[i].alloc_offset == WP_CONVENTIONAL)
+		if (zone_info[i].alloc_offset == WP_MISSING_DEV)
 			continue;
 
+		if (zone_info[i].alloc_offset == WP_CONVENTIONAL) {
+			u64 stripe_nr, full_stripe_nr;
+			u64 stripe_offset;
+			int stripe_index;
+
+			stripe_nr = div64_u64(last_alloc, map->stripe_size);
+			stripe_offset = stripe_nr * map->stripe_size;
+			full_stripe_nr = div_u64(stripe_nr, map->num_stripes);
+			div_u64_rem(stripe_nr, map->num_stripes, &stripe_index);
+
+			zone_info[i].alloc_offset =
+				full_stripe_nr * map->stripe_size;
+
+			if (stripe_index > i)
+				zone_info[i].alloc_offset += map->stripe_size;
+			else if (stripe_index == i)
+				zone_info[i].alloc_offset +=
+					(last_alloc - stripe_offset);
+		}
+
 		if (test_bit(0, active) != test_bit(i, active)) {
 			if (!btrfs_zone_activate(bg))
 				return -EIO;
@@ -1508,7 +1557,8 @@ static int btrfs_load_block_group_raid0(struct btrfs_block_group *bg,
 static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg,
 					 struct btrfs_chunk_map *map,
 					 struct zone_info *zone_info,
-					 unsigned long *active)
+					 unsigned long *active,
+					 u64 last_alloc)
 {
 	struct btrfs_fs_info *fs_info = bg->fs_info;
 
@@ -1519,8 +1569,7 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg,
 	}
 
 	for (int i = 0; i < map->num_stripes; i++) {
-		if (zone_info[i].alloc_offset == WP_MISSING_DEV ||
-		    zone_info[i].alloc_offset == WP_CONVENTIONAL)
+		if (zone_info[i].alloc_offset == WP_MISSING_DEV)
 			continue;
 
 		if (test_bit(0, active) != test_bit(i, active)) {
@@ -1531,6 +1580,29 @@ static int btrfs_load_block_group_raid10(struct btrfs_block_group *bg,
 				set_bit(BLOCK_GROUP_FLAG_ZONE_IS_ACTIVE, &bg->runtime_flags);
 		}
 
+		if (zone_info[i].alloc_offset == WP_CONVENTIONAL) {
+			u64 stripe_nr, full_stripe_nr;
+			u64 stripe_offset;
+			int stripe_index;
+
+			stripe_nr = div64_u64(last_alloc, map->stripe_size);
+			stripe_offset = stripe_nr * map->stripe_size;
+			full_stripe_nr = div_u64(stripe_nr,
+					 map->num_stripes / map->sub_stripes);
+			div_u64_rem(stripe_nr,
+				    (map->num_stripes / map->sub_stripes),
+				    &stripe_index);
+
+			zone_info[i].alloc_offset =
+				full_stripe_nr * map->stripe_size;
+
+			if (stripe_index > (i / map->sub_stripes))
+				zone_info[i].alloc_offset += map->stripe_size;
+			else if (stripe_index == (i / map->sub_stripes))
+				zone_info[i].alloc_offset +=
+					(last_alloc - stripe_offset);
+		}
+
 		if ((i % map->sub_stripes) == 0) {
 			bg->zone_capacity += zone_info[i].capacity;
 			bg->alloc_offset += zone_info[i].alloc_offset;
@@ -1584,7 +1656,7 @@ int btrfs_load_block_group_zone_info(struct btrfs_block_group *cache, bool new)
 	}
 
 	for (i = 0; i < map->num_stripes; i++) {
-		ret = btrfs_load_zone_info(fs_info, i, &zone_info[i], active, map);
+		ret = btrfs_load_zone_info(fs_info, i, &zone_info[i], active, map, new);
 		if (ret)
 			goto out;
 
@@ -1619,18 +1691,22 @@ int btrfs_load_block_group_zone_info(struct btrfs_block_group *cache, bool new)
 		ret = btrfs_load_block_group_single(cache, &zone_info[0], active);
 		break;
 	case BTRFS_BLOCK_GROUP_DUP:
-		ret = btrfs_load_block_group_dup(cache, map, zone_info, active);
+		ret = btrfs_load_block_group_dup(cache, map, zone_info, active,
+						 last_alloc);
 		break;
 	case BTRFS_BLOCK_GROUP_RAID1:
 	case BTRFS_BLOCK_GROUP_RAID1C3:
 	case BTRFS_BLOCK_GROUP_RAID1C4:
-		ret = btrfs_load_block_group_raid1(cache, map, zone_info, active);
+		ret = btrfs_load_block_group_raid1(cache, map, zone_info,
+						   active, last_alloc);
 		break;
 	case BTRFS_BLOCK_GROUP_RAID0:
-		ret = btrfs_load_block_group_raid0(cache, map, zone_info, active);
+		ret = btrfs_load_block_group_raid0(cache, map, zone_info,
+						   active, last_alloc);
 		break;
 	case BTRFS_BLOCK_GROUP_RAID10:
-		ret = btrfs_load_block_group_raid10(cache, map, zone_info, active);
+		ret = btrfs_load_block_group_raid10(cache, map, zone_info,
+						    active, last_alloc);
 		break;
 	case BTRFS_BLOCK_GROUP_RAID5:
 	case BTRFS_BLOCK_GROUP_RAID6:
@@ -1655,7 +1731,6 @@ int btrfs_load_block_group_zone_info(struct btrfs_block_group *cache, bool new)
 		 * stripe.
 		 */
 		cache->alloc_offset = cache->zone_capacity;
-		ret = 0;
 	}
 
 out:
@@ -1739,7 +1814,7 @@ bool btrfs_use_zone_append(struct btrfs_bio *bbio)
 		return false;
 
 	/*
-	 * Using REQ_OP_ZONE_APPNED for relocation can break assumptions on the
+	 * Using REQ_OP_ZONE_APPEND for relocation can break assumptions on the
 	 * extent layout the relocation code has.
 	 * Furthermore we have set aside own block-group from which only the
 	 * relocation "process" can allocate and make sure only one process at a
@@ -1780,12 +1855,12 @@ static void btrfs_rewrite_logical_zoned(struct btrfs_ordered_extent *ordered,
 	ordered->disk_bytenr = logical;
 
 	write_lock(&em_tree->lock);
-	em = search_extent_mapping(em_tree, ordered->file_offset,
-				   ordered->num_bytes);
+	em = btrfs_search_extent_mapping(em_tree, ordered->file_offset,
+					 ordered->num_bytes);
 	/* The em should be a new COW extent, thus it should not have an offset. */
 	ASSERT(em->offset == 0);
 	em->disk_bytenr = logical;
-	free_extent_map(em);
+	btrfs_free_extent_map(em);
 	write_unlock(&em_tree->lock);
 }
 
@@ -1795,8 +1870,8 @@ static bool btrfs_zoned_split_ordered(struct btrfs_ordered_extent *ordered,
 	struct btrfs_ordered_extent *new;
 
 	if (!test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags) &&
-	    split_extent_map(ordered->inode, ordered->file_offset,
-			     ordered->num_bytes, len, logical))
+	    btrfs_split_extent_map(ordered->inode, ordered->file_offset,
+				   ordered->num_bytes, len, logical))
 		return false;
 
 	new = btrfs_split_ordered_extent(ordered, len);
@@ -1973,7 +2048,7 @@ int btrfs_check_meta_write_pointer(struct btrfs_fs_info *fs_info,
 	if (block_group->meta_write_pointer > eb->start)
 		return -EBUSY;
 
-	/* If for_sync, this hole will be filled with trasnsaction commit. */
+	/* If for_sync, this hole will be filled with transaction commit. */
 	if (wbc->sync_mode == WB_SYNC_ALL && !wbc->for_sync)
 		return -EAGAIN;
 	return -EBUSY;
@@ -2107,6 +2182,9 @@ bool btrfs_zone_activate(struct btrfs_block_group *block_group)
 		physical = map->stripes[i].physical;
 		zinfo = device->zone_info;
 
+		if (!device->bdev)
+			continue;
+
 		if (zinfo->max_active_zones == 0)
 			continue;
 
@@ -2151,27 +2229,15 @@ static void wait_eb_writebacks(struct btrfs_block_group *block_group)
 {
 	struct btrfs_fs_info *fs_info = block_group->fs_info;
 	const u64 end = block_group->start + block_group->length;
-	struct radix_tree_iter iter;
 	struct extent_buffer *eb;
-	void __rcu **slot;
+	unsigned long index, start = (block_group->start >> fs_info->sectorsize_bits);
 
 	rcu_read_lock();
-	radix_tree_for_each_slot(slot, &fs_info->buffer_radix, &iter,
-				 block_group->start >> fs_info->sectorsize_bits) {
-		eb = radix_tree_deref_slot(slot);
-		if (!eb)
-			continue;
-		if (radix_tree_deref_retry(eb)) {
-			slot = radix_tree_iter_retry(&iter);
-			continue;
-		}
-
+	xa_for_each_start(&fs_info->buffer_tree, index, eb, start) {
 		if (eb->start < block_group->start)
 			continue;
 		if (eb->start >= end)
 			break;
-
-		slot = radix_tree_iter_resume(slot, &iter);
 		rcu_read_unlock();
 		wait_on_extent_buffer_writeback(eb);
 		rcu_read_lock();
@@ -2268,6 +2334,9 @@ static int do_zone_finish(struct btrfs_block_group *block_group, bool fully_writ
 		struct btrfs_zoned_device_info *zinfo = device->zone_info;
 		unsigned int nofs_flags;
 
+		if (!device->bdev)
+			continue;
+
 		if (zinfo->max_active_zones == 0)
 			continue;
 
@@ -2321,6 +2390,9 @@ bool btrfs_can_activate_zone(struct btrfs_fs_devices *fs_devices, u64 flags)
 	if (!btrfs_is_zoned(fs_info))
 		return true;
 
+	if (test_bit(BTRFS_FS_NEED_ZONE_FINISH, &fs_info->flags))
+		return false;
+
 	/* Check if there is a device with active zones left */
 	mutex_lock(&fs_info->chunk_mutex);
 	spin_lock(&fs_info->zone_active_bgs_lock);
@@ -2648,3 +2720,127 @@ void btrfs_check_active_zone_reservation(struct btrfs_fs_info *fs_info)
 	}
 	spin_unlock(&fs_info->zone_active_bgs_lock);
 }
+
+/*
+ * Reset the zones of unused block groups from @space_info->bytes_zone_unusable.
+ *
+ * @space_info:	the space to work on
+ * @num_bytes:	targeting reclaim bytes
+ *
+ * This one resets the zones of a block group, so we can reuse the region
+ * without removing the block group. On the other hand, btrfs_delete_unused_bgs()
+ * just removes a block group and frees up the underlying zones. So, we still
+ * need to allocate a new block group to reuse the zones.
+ *
+ * Resetting is faster than deleting/recreating a block group. It is similar
+ * to freeing the logical space on the regular mode. However, we cannot change
+ * the block group's profile with this operation.
+ */
+int btrfs_reset_unused_block_groups(struct btrfs_space_info *space_info, u64 num_bytes)
+{
+	struct btrfs_fs_info *fs_info = space_info->fs_info;
+	const sector_t zone_size_sectors = fs_info->zone_size >> SECTOR_SHIFT;
+
+	if (!btrfs_is_zoned(fs_info))
+		return 0;
+
+	while (num_bytes > 0) {
+		struct btrfs_chunk_map *map;
+		struct btrfs_block_group *bg = NULL;
+		bool found = false;
+		u64 reclaimed = 0;
+
+		/*
+		 * Here, we choose a fully zone_unusable block group. It's
+		 * technically possible to reset a partly zone_unusable block
+		 * group, which still has some free space left. However,
+		 * handling that needs to cope with the allocation side, which
+		 * makes the logic more complex. So, let's handle the easy case
+		 * for now.
+		 */
+		spin_lock(&fs_info->unused_bgs_lock);
+		list_for_each_entry(bg, &fs_info->unused_bgs, bg_list) {
+			if ((bg->flags & BTRFS_BLOCK_GROUP_TYPE_MASK) != space_info->flags)
+				continue;
+
+			/*
+			 * Use trylock to avoid locking order violation. In
+			 * btrfs_reclaim_bgs_work(), the lock order is
+			 * &bg->lock -> &fs_info->unused_bgs_lock. We skip a
+			 * block group if we cannot take its lock.
+			 */
+			if (!spin_trylock(&bg->lock))
+				continue;
+			if (btrfs_is_block_group_used(bg) || bg->zone_unusable < bg->length) {
+				spin_unlock(&bg->lock);
+				continue;
+			}
+			spin_unlock(&bg->lock);
+			found = true;
+			break;
+		}
+		if (!found) {
+			spin_unlock(&fs_info->unused_bgs_lock);
+			return 0;
+		}
+
+		list_del_init(&bg->bg_list);
+		btrfs_put_block_group(bg);
+		spin_unlock(&fs_info->unused_bgs_lock);
+
+		/*
+		 * Since the block group is fully zone_unusable and we cannot
+		 * allocate from this block group anymore, we don't need to set
+		 * this block group read-only.
+		 */
+
+		down_read(&fs_info->dev_replace.rwsem);
+		map = bg->physical_map;
+		for (int i = 0; i < map->num_stripes; i++) {
+			struct btrfs_io_stripe *stripe = &map->stripes[i];
+			unsigned int nofs_flags;
+			int ret;
+
+			nofs_flags = memalloc_nofs_save();
+			ret = blkdev_zone_mgmt(stripe->dev->bdev, REQ_OP_ZONE_RESET,
+					       stripe->physical >> SECTOR_SHIFT,
+					       zone_size_sectors);
+			memalloc_nofs_restore(nofs_flags);
+
+			if (ret) {
+				up_read(&fs_info->dev_replace.rwsem);
+				return ret;
+			}
+		}
+		up_read(&fs_info->dev_replace.rwsem);
+
+		spin_lock(&space_info->lock);
+		spin_lock(&bg->lock);
+		ASSERT(!btrfs_is_block_group_used(bg));
+		if (bg->ro) {
+			spin_unlock(&bg->lock);
+			spin_unlock(&space_info->lock);
+			continue;
+		}
+
+		reclaimed = bg->alloc_offset;
+		bg->zone_unusable = bg->length - bg->zone_capacity;
+		bg->alloc_offset = 0;
+		/*
+		 * This holds because we currently reset fully used then freed
+		 * block group.
+		 */
+		ASSERT(reclaimed == bg->zone_capacity);
+		bg->free_space_ctl->free_space += reclaimed;
+		space_info->bytes_zone_unusable -= reclaimed;
+		spin_unlock(&bg->lock);
+		btrfs_return_free_space(space_info, reclaimed);
+		spin_unlock(&space_info->lock);
+
+		if (num_bytes <= reclaimed)
+			break;
+		num_bytes -= reclaimed;
+	}
+
+	return 0;
+}