From: Josef Bacik <josef@toxicpanda.com>
Date: Tue, 21 Jul 2020 10:22:25 -0400
Subject: btrfs: add btrfs_reserve_data_bytes and use it
Git-commit: 8698fc4eb7884eeba29adc4193f9d05fa2bed16b
Patch-mainline: v5.10-rc1
References: bsc#1135481

Create a new function btrfs_reserve_data_bytes() in order to handle data
reservations.  This uses the new flush types and flush states to handle
making data reservations.

This patch specifically does not change any functionality, and is
purposefully not cleaned up in order to make bisection easier for the
future patches.  The new helper is identical to the old helper in how it
handles data reservations.  We first try to force a chunk allocation,
and then we run through the flush states all at once and in the same
order that they were done with the old helper.

Subsequent patches will clean this up and change the behavior of the
flushing, and it is important to keep those changes separate so we can
easily bisect down to the patch that caused the regression, rather than
the patch that made us start using the new infrastructure.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Tested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Acked-by: Nikolay Borisov <nborisov@suse.com>
---
 fs/btrfs/delalloc-space.c |  119 +---------------------------------------------
 fs/btrfs/space-info.c     |   98 +++++++++++++++++++++++++++++++++++++
 fs/btrfs/space-info.h     |    2 
 3 files changed, 104 insertions(+), 115 deletions(-)

--- a/fs/btrfs/delalloc-space.c
+++ b/fs/btrfs/delalloc-space.c
@@ -13,126 +13,15 @@ int btrfs_alloc_data_chunk_ondemand(stru
 {
 	struct btrfs_root *root = inode->root;
 	struct btrfs_fs_info *fs_info = root->fs_info;
-	struct btrfs_space_info *data_sinfo = fs_info->data_sinfo;
-	u64 used;
-	int ret = 0;
-	int need_commit = 2;
-	int have_pinned_space;
+	enum btrfs_reserve_flush_enum flush = BTRFS_RESERVE_FLUSH_DATA;
 
 	/* Make sure bytes are sectorsize aligned */
 	bytes = ALIGN(bytes, fs_info->sectorsize);
 
-	if (btrfs_is_free_space_inode(inode)) {
-		need_commit = 0;
-		ASSERT(current->journal_info);
-	}
+	if (btrfs_is_free_space_inode(inode))
+		flush = BTRFS_RESERVE_FLUSH_FREE_SPACE_INODE;
 
-again:
-	/* Make sure we have enough space to handle the data first */
-	spin_lock(&data_sinfo->lock);
-	used = btrfs_space_info_used(data_sinfo, true);
-
-	if (used + bytes > data_sinfo->total_bytes) {
-		struct btrfs_trans_handle *trans;
-
-		/*
-		 * If we don't have enough free bytes in this space then we need
-		 * to alloc a new chunk.
-		 */
-		if (!data_sinfo->full) {
-			u64 alloc_target;
-
-			data_sinfo->force_alloc = CHUNK_ALLOC_FORCE;
-			spin_unlock(&data_sinfo->lock);
-
-			alloc_target = btrfs_data_alloc_profile(fs_info);
-			/*
-			 * It is ugly that we don't call nolock join
-			 * transaction for the free space inode case here.
-			 * But it is safe because we only do the data space
-			 * reservation for the free space cache in the
-			 * transaction context, the common join transaction
-			 * just increase the counter of the current transaction
-			 * handler, doesn't try to acquire the trans_lock of
-			 * the fs.
-			 */
-			trans = btrfs_join_transaction(root);
-			if (IS_ERR(trans))
-				return PTR_ERR(trans);
-
-			ret = btrfs_chunk_alloc(trans, alloc_target,
-						CHUNK_ALLOC_NO_FORCE);
-			btrfs_end_transaction(trans);
-			if (ret < 0) {
-				if (ret != -ENOSPC)
-					return ret;
-				else {
-					have_pinned_space = 1;
-					goto commit_trans;
-				}
-			}
-
-			goto again;
-		}
-
-		/*
-		 * If we don't have enough pinned space to deal with this
-		 * allocation, and no removed chunk in current transaction,
-		 * don't bother committing the transaction.
-		 */
-		have_pinned_space = __percpu_counter_compare(
-			&data_sinfo->total_bytes_pinned,
-			used + bytes - data_sinfo->total_bytes,
-			BTRFS_TOTAL_BYTES_PINNED_BATCH);
-		spin_unlock(&data_sinfo->lock);
-
-		/* Commit the current transaction and try again */
-commit_trans:
-		if (need_commit && !atomic_read(&fs_info->open_ioctl_trans)) {
-			need_commit--;
-
-			if (need_commit > 0) {
-				btrfs_start_delalloc_roots(fs_info, U64_MAX);
-				btrfs_wait_ordered_roots(fs_info, U64_MAX, 0,
-							 (u64)-1);
-			}
-
-			trans = btrfs_join_transaction(root);
-			if (IS_ERR(trans))
-				return PTR_ERR(trans);
-			if (have_pinned_space >= 0 ||
-			    test_bit(BTRFS_TRANS_HAVE_FREE_BGS,
-				     &trans->transaction->flags) ||
-			    need_commit > 0) {
-				ret = btrfs_commit_transaction(trans);
-				if (ret)
-					return ret;
-				/*
-				 * The cleaner kthread might still be doing iput
-				 * operations. Wait for it to finish so that
-				 * more space is released.  We don't need to
-				 * explicitly run the delayed iputs here because
-				 * the commit_transaction would have woken up
-				 * the cleaner.
-				 */
-				ret = btrfs_wait_on_delayed_iputs(fs_info);
-				if (ret)
-					return ret;
-				goto again;
-			} else {
-				btrfs_end_transaction(trans);
-			}
-		}
-
-		trace_btrfs_space_reservation(fs_info,
-					      "space_info:enospc",
-					      data_sinfo->flags, bytes, 1);
-		return -ENOSPC;
-	}
-	btrfs_space_info_update_bytes_may_use(fs_info, data_sinfo, bytes);
-	spin_unlock(&data_sinfo->lock);
-
-	return ret;
+	return btrfs_reserve_data_bytes(fs_info, bytes, flush);
 }
 
 int btrfs_check_data_free_space(struct inode *inode,
--- a/fs/btrfs/space-info.c
+++ b/fs/btrfs/space-info.c
@@ -1207,3 +1207,101 @@ int btrfs_reserve_metadata_bytes(struct
 	return ret;
 }
 
+/**
+ * btrfs_reserve_data_bytes - try to reserve data bytes for an allocation
+ * @fs_info - the filesystem
+ * @bytes - the number of bytes we need
+ * @flush - how we are allowed to flush
+ *
+ * This will reserve bytes from the data space info.  If there is not enough
+ * space then we will attempt to flush space as specified by flush.
+ */
+int btrfs_reserve_data_bytes(struct btrfs_fs_info *fs_info, u64 bytes,
+			     enum btrfs_reserve_flush_enum flush)
+{
+	struct btrfs_space_info *data_sinfo = fs_info->data_sinfo;
+	const enum btrfs_flush_state *states = NULL;
+	u64 used;
+	int states_nr = 0;
+	int commit_cycles = 2;
+	int ret = -ENOSPC;
+
+	ASSERT(!current->journal_info || flush != BTRFS_RESERVE_FLUSH_DATA);
+
+	if (flush == BTRFS_RESERVE_FLUSH_DATA) {
+		states = data_flush_states;
+		states_nr = ARRAY_SIZE(data_flush_states);
+	}
+
+	spin_lock(&data_sinfo->lock);
+again:
+	used = btrfs_space_info_used(data_sinfo, true);
+
+	if (used + bytes > data_sinfo->total_bytes) {
+		u64 prev_total_bytes = data_sinfo->total_bytes;
+		int flush_state = 0;
+
+		spin_unlock(&data_sinfo->lock);
+
+		/*
+		 * Everybody can force chunk allocation, so try this first to
+		 * see if we can just bail here and make our reservation.
+		 */
+		flush_space(fs_info, data_sinfo, bytes, ALLOC_CHUNK_FORCE);
+		spin_lock(&data_sinfo->lock);
+		if (prev_total_bytes < data_sinfo->total_bytes)
+			goto again;
+		spin_unlock(&data_sinfo->lock);
+
+		/*
+		 * Cycle through the rest of the flushing options for our flush
+		 * type, then try again.
+		 */
+		while (flush_state < states_nr) {
+			u64 flush_bytes = U64_MAX;
+
+			/*
+			 * Previously we unconditionally committed the
+			 * transaction twice before finally checking against
+			 * pinned space before committing the final time.  We
+			 * also skipped flushing delalloc the final pass
+			 * through.
+			 */
+			if (!commit_cycles) {
+				if (states[flush_state] == FLUSH_DELALLOC_WAIT) {
+					flush_state++;
+					continue;
+				}
+				if (states[flush_state] == COMMIT_TRANS)
+					flush_bytes = bytes;
+			}
+
+			if (states[flush_state] == COMMIT_TRANS &&
+			    atomic_read(&fs_info->open_ioctl_trans)) {
+				flush_state++;
+				continue;
+			}
+
+			flush_space(fs_info, data_sinfo, flush_bytes,
+				    states[flush_state]);
+			flush_state++;
+		}
+
+		if (!commit_cycles)
+			goto out;
+
+		commit_cycles--;
+		spin_lock(&data_sinfo->lock);
+		goto again;
+	}
+	btrfs_space_info_update_bytes_may_use(fs_info, data_sinfo, bytes);
+	ret = 0;
+	spin_unlock(&data_sinfo->lock);
+out:
+	if (ret)
+		trace_btrfs_space_reservation(fs_info,
+					      "space_info:enospc",
+					      data_sinfo->flags, bytes, 1);
+	return ret;
+}
+
--- a/fs/btrfs/space-info.h
+++ b/fs/btrfs/space-info.h
@@ -150,5 +150,7 @@ static inline void btrfs_space_info_free
 	btrfs_try_granting_tickets(fs_info, space_info);
 	spin_unlock(&space_info->lock);
 }
+int btrfs_reserve_data_bytes(struct btrfs_fs_info *fs_info, u64 bytes,
+			     enum btrfs_reserve_flush_enum flush);
 
 #endif /* BTRFS_SPACE_INFO_H */