btrfs: use an efficient way to represent source of duplicated stripes

For btrfs dev-replace, we have to duplicate writes to the source
device into the target device.

For non-RAID56, all writes into the same mapped ranges are sharing the
same content, thus they don't really need to bother anything.
(E.g. in btrfs_submit_bio() for non-RAID56 range we just submit the
same write to all involved devices).

But for RAID56, all stripes contain different content, thus we must
have a clear mapping of which stripe is duplicated from which original
stripe.

Currently we use a complex way using tgtdev_map[] array, e.g:

 num_tgtdevs = 1
 tgtdev_map[0] = 0    <- Means stripes[0] is not involved in replace.
 tgtdev_map[1] = 3    <- Means stripes[1] is involved in replace,
			 and it's duplicated to stripes[3].
 tgtdev_map[2] = 0    <- Means stripes[2] is not involved in replace.

But this is wasting some space, and ignores one important thing for
dev-replace, there is at most one running replace.

Thus we can change it to a fixed array to represent the mapping:

 replace_nr_stripes = 1
 replace_stripe_src = 1    <- Means stripes[1] is involved in replace.
			      thus the extra stripe is a copy of
			      stripes[1]

By this we can save some space for bioc on RAID56 chunks with many
devices.  And we get rid of one variable sized array from bioc.

Thus the patch involves the following changes:

- Replace @num_tgtdevs and @tgtdev_map[] with @replace_nr_stripes
  and @replace_stripe_src.

  @num_tgtdevs is just renamed to @replace_nr_stripes.
  While the mapping is completely changed.

- Add extra ASSERT()s for RAID56 code

- Only add two more extra stripes for dev-replace cases.
  As we have an upper limit on how many dev-replace stripes we can have.

- Unify the behavior of handle_ops_on_dev_replace()
  Previously handle_ops_on_dev_replace() go two different paths for
  WRITE and GET_READ_MIRRORS.
  Now unify them by always going the WRITE path first (with at most 2
  replace stripes), then if we're doing GET_READ_MIRRORS and we have 2
  extra stripes, just drop one stripe.

- Remove the @real_stripes argument from alloc_btrfs_io_context()
  As we don't need the old variable length array any more.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

1 files changed, 71 insertions, 91 deletions

diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 2417f4fb8724..8f06f0e47ba8 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -5914,8 +5914,7 @@ static void sort_parity_stripes(struct btrfs_io_context *bioc, int num_stripes)
 }
 
 static struct btrfs_io_context *alloc_btrfs_io_context(struct btrfs_fs_info *fs_info,
-						       u16 total_stripes,
-						       u16 real_stripes)
+						       u16 total_stripes)
 {
 	struct btrfs_io_context *bioc;
 
@@ -5924,8 +5923,6 @@ static struct btrfs_io_context *alloc_btrfs_io_context(struct btrfs_fs_info *fs_
 		sizeof(struct btrfs_io_context) +
 		/* Plus the variable array for the stripes */
 		sizeof(struct btrfs_io_stripe) * (total_stripes) +
-		/* Plus the variable array for the tgt dev */
-		sizeof(u16) * (real_stripes) +
 		/*
 		 * Plus the raid_map, which includes both the tgt dev
 		 * and the stripes.
@@ -5939,8 +5936,8 @@ static struct btrfs_io_context *alloc_btrfs_io_context(struct btrfs_fs_info *fs_
 	refcount_set(&bioc->refs, 1);
 
 	bioc->fs_info = fs_info;
-	bioc->tgtdev_map = (u16 *)(bioc->stripes + total_stripes);
-	bioc->raid_map = (u64 *)(bioc->tgtdev_map + real_stripes);
+	bioc->raid_map = (u64 *)(bioc->stripes + total_stripes);
+	bioc->replace_stripe_src = -1;
 
 	return bioc;
 }
@@ -6204,93 +6201,74 @@ static void handle_ops_on_dev_replace(enum btrfs_map_op op,
 				      int *num_stripes_ret, int *max_errors_ret)
 {
 	u64 srcdev_devid = dev_replace->srcdev->devid;
-	int tgtdev_indexes = 0;
+	/*
+	 * At this stage, num_stripes is still the real number of stripes,
+	 * excluding the duplicated stripes.
+	 */
 	int num_stripes = *num_stripes_ret;
+	int nr_extra_stripes = 0;
 	int max_errors = *max_errors_ret;
 	int i;
 
-	if (op == BTRFS_MAP_WRITE) {
-		int index_where_to_add;
+	/*
+	 * A block group which has "to_copy" set will eventually be copied by
+	 * the dev-replace process. We can avoid cloning IO here.
+	 */
+	if (is_block_group_to_copy(dev_replace->srcdev->fs_info, logical))
+		return;
 
-		/*
-		 * A block group which have "to_copy" set will eventually
-		 * copied by dev-replace process. We can avoid cloning IO here.
-		 */
-		if (is_block_group_to_copy(dev_replace->srcdev->fs_info, logical))
-			return;
+	/*
+	 * Duplicate the write operations while the dev-replace procedure is
+	 * running. Since the copying of the old disk to the new disk takes
+	 * place at run time while the filesystem is mounted writable, the
+	 * regular write operations to the old disk have to be duplicated to go
+	 * to the new disk as well.
+	 *
+	 * Note that device->missing is handled by the caller, and that the
+	 * write to the old disk is already set up in the stripes array.
+	 */
+	for (i = 0; i < num_stripes; i++) {
+		struct btrfs_io_stripe *old = &bioc->stripes[i];
+		struct btrfs_io_stripe *new = &bioc->stripes[num_stripes + nr_extra_stripes];
 
-		/*
-		 * duplicate the write operations while the dev replace
-		 * procedure is running. Since the copying of the old disk to
-		 * the new disk takes place at run time while the filesystem is
-		 * mounted writable, the regular write operations to the old
-		 * disk have to be duplicated to go to the new disk as well.
-		 *
-		 * Note that device->missing is handled by the caller, and that
-		 * the write to the old disk is already set up in the stripes
-		 * array.
-		 */
-		index_where_to_add = num_stripes;
-		for (i = 0; i < num_stripes; i++) {
-			if (bioc->stripes[i].dev->devid == srcdev_devid) {
-				/* write to new disk, too */
-				struct btrfs_io_stripe *new =
-					bioc->stripes + index_where_to_add;
-				struct btrfs_io_stripe *old =
-					bioc->stripes + i;
-
-				new->physical = old->physical;
-				new->dev = dev_replace->tgtdev;
-				bioc->tgtdev_map[i] = index_where_to_add;
-				index_where_to_add++;
-				max_errors++;
-				tgtdev_indexes++;
-			}
-		}
-		num_stripes = index_where_to_add;
-	} else if (op == BTRFS_MAP_GET_READ_MIRRORS) {
-		int index_srcdev = 0;
-		int found = 0;
-		u64 physical_of_found = 0;
+		if (old->dev->devid != srcdev_devid)
+			continue;
 
-		/*
-		 * During the dev-replace procedure, the target drive can also
-		 * be used to read data in case it is needed to repair a corrupt
-		 * block elsewhere. This is possible if the requested area is
-		 * left of the left cursor. In this area, the target drive is a
-		 * full copy of the source drive.
-		 */
-		for (i = 0; i < num_stripes; i++) {
-			if (bioc->stripes[i].dev->devid == srcdev_devid) {
-				/*
-				 * In case of DUP, in order to keep it simple,
-				 * only add the mirror with the lowest physical
-				 * address
-				 */
-				if (found &&
-				    physical_of_found <= bioc->stripes[i].physical)
-					continue;
-				index_srcdev = i;
-				found = 1;
-				physical_of_found = bioc->stripes[i].physical;
-			}
-		}
-		if (found) {
-			struct btrfs_io_stripe *tgtdev_stripe =
-				bioc->stripes + num_stripes;
+		new->physical = old->physical;
+		new->dev = dev_replace->tgtdev;
+		if (bioc->map_type & BTRFS_BLOCK_GROUP_RAID56_MASK)
+			bioc->replace_stripe_src = i;
+		nr_extra_stripes++;
+	}
 
-			tgtdev_stripe->physical = physical_of_found;
-			tgtdev_stripe->dev = dev_replace->tgtdev;
-			bioc->tgtdev_map[index_srcdev] = num_stripes;
+	/* We can only have at most 2 extra nr_stripes (for DUP). */
+	ASSERT(nr_extra_stripes <= 2);
+	/*
+	 * For GET_READ_MIRRORS, we can only return at most 1 extra stripe for
+	 * replace.
+	 * If we have 2 extra stripes, only choose the one with smaller physical.
+	 */
+	if (op == BTRFS_MAP_GET_READ_MIRRORS && nr_extra_stripes == 2) {
+		struct btrfs_io_stripe *first = &bioc->stripes[num_stripes];
+		struct btrfs_io_stripe *second = &bioc->stripes[num_stripes + 1];
 
-			tgtdev_indexes++;
-			num_stripes++;
+		/* Only DUP can have two extra stripes. */
+		ASSERT(bioc->map_type & BTRFS_BLOCK_GROUP_DUP);
+
+		/*
+		 * Swap the last stripe stripes and reduce @nr_extra_stripes.
+		 * The extra stripe would still be there, but won't be accessed.
+		 */
+		if (first->physical > second->physical) {
+			swap(second->physical, first->physical);
+			swap(second->dev, first->dev);
+			nr_extra_stripes--;
 		}
 	}
 
-	*num_stripes_ret = num_stripes;
-	*max_errors_ret = max_errors;
-	bioc->num_tgtdevs = tgtdev_indexes;
+	*num_stripes_ret = num_stripes + nr_extra_stripes;
+	*max_errors_ret = max_errors + nr_extra_stripes;
+	bioc->replace_nr_stripes = nr_extra_stripes;
 }
 
 static bool need_full_stripe(enum btrfs_map_op op)
@@ -6377,7 +6355,6 @@ int __btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
 	int dev_replace_is_ongoing = 0;
 	int patch_the_first_stripe_for_dev_replace = 0;
 	u16 num_alloc_stripes;
-	u16 tgtdev_indexes = 0;
 	u64 physical_to_patch_in_first_stripe = 0;
 	u64 raid56_full_stripe_start = (u64)-1;
 	u64 max_len;
@@ -6523,13 +6500,16 @@ int __btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
 	}
 
 	num_alloc_stripes = num_stripes;
-	if (dev_replace_is_ongoing && dev_replace->tgtdev != NULL) {
-		if (op == BTRFS_MAP_WRITE)
-			num_alloc_stripes <<= 1;
-		if (op == BTRFS_MAP_GET_READ_MIRRORS)
-			num_alloc_stripes++;
-		tgtdev_indexes = num_stripes;
-	}
+	if (dev_replace_is_ongoing && dev_replace->tgtdev != NULL &&
+	    op != BTRFS_MAP_READ)
+		/*
+		 * For replace case, we need to add extra stripes for extra
+		 * duplicated stripes.
+		 *
+		 * For both WRITE and GET_READ_MIRRORS, we may have at most
+		 * 2 more stripes (DUP types, otherwise 1).
+		 */
+		num_alloc_stripes += 2;
 
 	/*
 	 * If this I/O maps to a single device, try to return the device and
@@ -6554,11 +6534,12 @@ int __btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
 		goto out;
 	}
 
-	bioc = alloc_btrfs_io_context(fs_info, num_alloc_stripes, tgtdev_indexes);
+	bioc = alloc_btrfs_io_context(fs_info, num_alloc_stripes);
 	if (!bioc) {
 		ret = -ENOMEM;
 		goto out;
 	}
+	bioc->map_type = map->type;
 
 	for (i = 0; i < num_stripes; i++) {
 		set_io_stripe(&bioc->stripes[i], map, stripe_index, stripe_offset,
@@ -6599,7 +6580,6 @@ int __btrfs_map_block(struct btrfs_fs_info *fs_info, enum btrfs_map_op op,
 	}
 
 	*bioc_ret = bioc;
-	bioc->map_type = map->type;
 	bioc->num_stripes = num_stripes;
 	bioc->max_errors = max_errors;
 	bioc->mirror_num = mirror_num;