Btrfs: rework outstanding_extents

Right now we do a lot of weird hoops around outstanding_extents in order
to keep the extent count consistent.  This is because we logically
transfer the outstanding_extent count from the initial reservation
through the set_delalloc_bits.  This makes it pretty difficult to get a
handle on how and when we need to mess with outstanding_extents.

Fix this by revamping the rules of how we deal with outstanding_extents.
Now instead everybody that is holding on to a delalloc extent is
required to increase the outstanding extents count for itself.  This
means we'll have something like this

btrfs_delalloc_reserve_metadata	- outstanding_extents = 1
 btrfs_set_extent_delalloc	- outstanding_extents = 2
btrfs_release_delalloc_extents	- outstanding_extents = 1

for an initial file write.  Now take the append write where we extend an
existing delalloc range but still under the maximum extent size

btrfs_delalloc_reserve_metadata - outstanding_extents = 2
  btrfs_set_extent_delalloc
    btrfs_set_bit_hook		- outstanding_extents = 3
    btrfs_merge_extent_hook	- outstanding_extents = 2
btrfs_delalloc_release_extents	- outstanding_extnets = 1

In order to make the ordered extent transition we of course must now
make ordered extents carry their own outstanding_extent reservation, so
for cow_file_range we end up with

btrfs_add_ordered_extent	- outstanding_extents = 2
clear_extent_bit		- outstanding_extents = 1
btrfs_remove_ordered_extent	- outstanding_extents = 0

This makes all manipulations of outstanding_extents much more explicit.
Every successful call to btrfs_delalloc_reserve_metadata _must_ now be
combined with btrfs_release_delalloc_extents, even in the error case, as
that is the only function that actually modifies the
outstanding_extents counter.

The drawback to this is now we are much more likely to have transient
cases where outstanding_extents is much larger than it actually should
be.  This could happen before as we manipulated the delalloc bits, but
now it happens basically at every write.  This may put more pressure on
the ENOSPC flushing code, but I think making this code simpler is worth
the cost.  I have another change coming to mitigate this side-effect
somewhat.

I also added trace points for the counter manipulation.  These were used
by a bpf script I wrote to help track down leak issues.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

1 files changed, 36 insertions, 78 deletions

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index eb20239284a2..5b0de1e120d1 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -67,7 +67,6 @@ struct btrfs_iget_args {
 };
 
 struct btrfs_dio_data {
-	u64 outstanding_extents;
 	u64 reserve;
 	u64 unsubmitted_oe_range_start;
 	u64 unsubmitted_oe_range_end;
@@ -348,7 +347,6 @@ static noinline int cow_file_range_inline(struct btrfs_root *root,
 	}
 
 	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
-	btrfs_delalloc_release_metadata(BTRFS_I(inode), end + 1 - start);
 	btrfs_drop_extent_cache(BTRFS_I(inode), start, aligned_end - 1, 0);
 out:
 	/*
@@ -584,16 +582,21 @@ cont:
 		}
 		if (ret <= 0) {
 			unsigned long clear_flags = EXTENT_DELALLOC |
-				EXTENT_DELALLOC_NEW | EXTENT_DEFRAG;
+				EXTENT_DELALLOC_NEW | EXTENT_DEFRAG |
+				EXTENT_DO_ACCOUNTING;
 			unsigned long page_error_op;
 
-			clear_flags |= (ret < 0) ? EXTENT_DO_ACCOUNTING : 0;
 			page_error_op = ret < 0 ? PAGE_SET_ERROR : 0;
 
 			/*
 			 * inline extent creation worked or returned error,
 			 * we don't need to create any more async work items.
 			 * Unlock and free up our temp pages.
+			 *
+			 * We use DO_ACCOUNTING here because we need the
+			 * delalloc_release_metadata to be done _after_ we drop
+			 * our outstanding extent for clearing delalloc for this
+			 * range.
 			 */
 			extent_clear_unlock_delalloc(inode, start, end, end,
 						     NULL, clear_flags,
@@ -602,10 +605,6 @@ cont:
 						     PAGE_SET_WRITEBACK |
 						     page_error_op |
 						     PAGE_END_WRITEBACK);
-			if (ret == 0)
-				btrfs_free_reserved_data_space_noquota(inode,
-							       start,
-							       end - start + 1);
 			goto free_pages_out;
 		}
 	}
@@ -981,15 +980,19 @@ static noinline int cow_file_range(struct inode *inode,
 		ret = cow_file_range_inline(root, inode, start, end, 0,
 					BTRFS_COMPRESS_NONE, NULL);
 		if (ret == 0) {
+			/*
+			 * We use DO_ACCOUNTING here because we need the
+			 * delalloc_release_metadata to be run _after_ we drop
+			 * our outstanding extent for clearing delalloc for this
+			 * range.
+			 */
 			extent_clear_unlock_delalloc(inode, start, end,
 				     delalloc_end, NULL,
 				     EXTENT_LOCKED | EXTENT_DELALLOC |
-				     EXTENT_DELALLOC_NEW |
-				     EXTENT_DEFRAG, PAGE_UNLOCK |
+				     EXTENT_DELALLOC_NEW | EXTENT_DEFRAG |
+				     EXTENT_DO_ACCOUNTING, PAGE_UNLOCK |
 				     PAGE_CLEAR_DIRTY | PAGE_SET_WRITEBACK |
 				     PAGE_END_WRITEBACK);
-			btrfs_free_reserved_data_space_noquota(inode, start,
-						end - start + 1);
 			*nr_written = *nr_written +
 			     (end - start + PAGE_SIZE) / PAGE_SIZE;
 			*page_started = 1;
@@ -1627,7 +1630,7 @@ static void btrfs_split_extent_hook(void *private_data,
 	}
 
 	spin_lock(&BTRFS_I(inode)->lock);
-	BTRFS_I(inode)->outstanding_extents++;
+	btrfs_mod_outstanding_extents(BTRFS_I(inode), 1);
 	spin_unlock(&BTRFS_I(inode)->lock);
 }
 
@@ -1657,7 +1660,7 @@ static void btrfs_merge_extent_hook(void *private_data,
 	/* we're not bigger than the max, unreserve the space and go */
 	if (new_size <= BTRFS_MAX_EXTENT_SIZE) {
 		spin_lock(&BTRFS_I(inode)->lock);
-		BTRFS_I(inode)->outstanding_extents--;
+		btrfs_mod_outstanding_extents(BTRFS_I(inode), -1);
 		spin_unlock(&BTRFS_I(inode)->lock);
 		return;
 	}
@@ -1688,7 +1691,7 @@ static void btrfs_merge_extent_hook(void *private_data,
 		return;
 
 	spin_lock(&BTRFS_I(inode)->lock);
-	BTRFS_I(inode)->outstanding_extents--;
+	btrfs_mod_outstanding_extents(BTRFS_I(inode), -1);
 	spin_unlock(&BTRFS_I(inode)->lock);
 }
 
@@ -1758,15 +1761,12 @@ static void btrfs_set_bit_hook(void *private_data,
 	if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
 		struct btrfs_root *root = BTRFS_I(inode)->root;
 		u64 len = state->end + 1 - state->start;
+		u32 num_extents = count_max_extents(len);
 		bool do_list = !btrfs_is_free_space_inode(BTRFS_I(inode));
 
-		if (*bits & EXTENT_FIRST_DELALLOC) {
-			*bits &= ~EXTENT_FIRST_DELALLOC;
-		} else {
-			spin_lock(&BTRFS_I(inode)->lock);
-			BTRFS_I(inode)->outstanding_extents++;
-			spin_unlock(&BTRFS_I(inode)->lock);
-		}
+		spin_lock(&BTRFS_I(inode)->lock);
+		btrfs_mod_outstanding_extents(BTRFS_I(inode), num_extents);
+		spin_unlock(&BTRFS_I(inode)->lock);
 
 		/* For sanity tests */
 		if (btrfs_is_testing(fs_info))
@@ -1820,13 +1820,9 @@ static void btrfs_clear_bit_hook(void *private_data,
 		struct btrfs_root *root = inode->root;
 		bool do_list = !btrfs_is_free_space_inode(inode);
 
-		if (*bits & EXTENT_FIRST_DELALLOC) {
-			*bits &= ~EXTENT_FIRST_DELALLOC;
-		} else if (!(*bits & EXTENT_CLEAR_META_RESV)) {
-			spin_lock(&inode->lock);
-			inode->outstanding_extents -= num_extents;
-			spin_unlock(&inode->lock);
-		}
+		spin_lock(&inode->lock);
+		btrfs_mod_outstanding_extents(inode, -num_extents);
+		spin_unlock(&inode->lock);
 
 		/*
 		 * We don't reserve metadata space for space cache inodes so we
@@ -2097,6 +2093,7 @@ again:
 				  0);
 	ClearPageChecked(page);
 	set_page_dirty(page);
+	btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
 out:
 	unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end,
 			     &cached_state, GFP_NOFS);
@@ -3051,9 +3048,6 @@ out:
 				 0, &cached_state, GFP_NOFS);
 	}
 
-	if (root != fs_info->tree_root)
-		btrfs_delalloc_release_metadata(BTRFS_I(inode),
-				ordered_extent->len);
 	if (trans)
 		btrfs_end_transaction(trans);
 
@@ -4794,8 +4788,11 @@ int btrfs_truncate_block(struct inode *inode, loff_t from, loff_t len,
 	    (!len || ((len & (blocksize - 1)) == 0)))
 		goto out;
 
+	block_start = round_down(from, blocksize);
+	block_end = block_start + blocksize - 1;
+
 	ret = btrfs_delalloc_reserve_space(inode, &data_reserved,
-			round_down(from, blocksize), blocksize);
+					   block_start, blocksize);
 	if (ret)
 		goto out;
 
@@ -4803,15 +4800,12 @@ again:
 	page = find_or_create_page(mapping, index, mask);
 	if (!page) {
 		btrfs_delalloc_release_space(inode, data_reserved,
-				round_down(from, blocksize),
-				blocksize);
+					     block_start, blocksize);
+		btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize);
 		ret = -ENOMEM;
 		goto out;
 	}
 
-	block_start = round_down(from, blocksize);
-	block_end = block_start + blocksize - 1;
-
 	if (!PageUptodate(page)) {
 		ret = btrfs_readpage(NULL, page);
 		lock_page(page);
@@ -4876,6 +4870,7 @@ out_unlock:
 	if (ret)
 		btrfs_delalloc_release_space(inode, data_reserved, block_start,
 					     blocksize);
+	btrfs_delalloc_release_extents(BTRFS_I(inode), blocksize);
 	unlock_page(page);
 	put_page(page);
 out:
@@ -7790,33 +7785,6 @@ static struct extent_map *create_io_em(struct inode *inode, u64 start, u64 len,
 	return em;
 }
 
-static void adjust_dio_outstanding_extents(struct inode *inode,
-					   struct btrfs_dio_data *dio_data,
-					   const u64 len)
-{
-	unsigned num_extents = count_max_extents(len);
-
-	/*
-	 * If we have an outstanding_extents count still set then we're
-	 * within our reservation, otherwise we need to adjust our inode
-	 * counter appropriately.
-	 */
-	if (dio_data->outstanding_extents >= num_extents) {
-		dio_data->outstanding_extents -= num_extents;
-	} else {
-		/*
-		 * If dio write length has been split due to no large enough
-		 * contiguous space, we need to compensate our inode counter
-		 * appropriately.
-		 */
-		u64 num_needed = num_extents - dio_data->outstanding_extents;
-
-		spin_lock(&BTRFS_I(inode)->lock);
-		BTRFS_I(inode)->outstanding_extents += num_needed;
-		spin_unlock(&BTRFS_I(inode)->lock);
-	}
-}
-
 static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
 				   struct buffer_head *bh_result, int create)
 {
@@ -7978,7 +7946,6 @@ unlock:
 		if (!dio_data->overwrite && start + len > i_size_read(inode))
 			i_size_write(inode, start + len);
 
-		adjust_dio_outstanding_extents(inode, dio_data, len);
 		WARN_ON(dio_data->reserve < len);
 		dio_data->reserve -= len;
 		dio_data->unsubmitted_oe_range_end = start + len;
@@ -8008,14 +7975,6 @@ unlock_err:
 err:
 	if (dio_data)
 		current->journal_info = dio_data;
-	/*
-	 * Compensate the delalloc release we do in btrfs_direct_IO() when we
-	 * write less data then expected, so that we don't underflow our inode's
-	 * outstanding extents counter.
-	 */
-	if (create && dio_data)
-		adjust_dio_outstanding_extents(inode, dio_data, len);
-
 	return ret;
 }
 
@@ -8860,7 +8819,6 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
 						   offset, count);
 		if (ret)
 			goto out;
-		dio_data.outstanding_extents = count_max_extents(count);
 
 		/*
 		 * We need to know how many extents we reserved so that we can
@@ -8887,6 +8845,7 @@ static ssize_t btrfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
 	if (iov_iter_rw(iter) == WRITE) {
 		up_read(&BTRFS_I(inode)->dio_sem);
 		current->journal_info = NULL;
+		btrfs_delalloc_release_extents(BTRFS_I(inode), count);
 		if (ret < 0 && ret != -EIOCBQUEUED) {
 			if (dio_data.reserve)
 				btrfs_delalloc_release_space(inode, data_reserved,
@@ -9224,9 +9183,6 @@ again:
 					  fs_info->sectorsize);
 		if (reserved_space < PAGE_SIZE) {
 			end = page_start + reserved_space - 1;
-			spin_lock(&BTRFS_I(inode)->lock);
-			BTRFS_I(inode)->outstanding_extents++;
-			spin_unlock(&BTRFS_I(inode)->lock);
 			btrfs_delalloc_release_space(inode, data_reserved,
 					page_start, PAGE_SIZE - reserved_space);
 		}
@@ -9278,12 +9234,14 @@ again:
 
 out_unlock:
 	if (!ret) {
+		btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
 		sb_end_pagefault(inode->i_sb);
 		extent_changeset_free(data_reserved);
 		return VM_FAULT_LOCKED;
 	}
 	unlock_page(page);
 out:
+	btrfs_delalloc_release_extents(BTRFS_I(inode), PAGE_SIZE);
 	btrfs_delalloc_release_space(inode, data_reserved, page_start,
 				     reserved_space);
 out_noreserve: