Btrfs: release delalloc reservations on extent item insertion

This patch fixes an issue with the delalloc metadata space reservation
code.  The problem is we used to free the reservation as soon as we
allocated the delalloc region.  The problem with this is if we are not
inserting an inline extent, we don't actually insert the extent item until
after the ordered extent is written out.  This patch does 3 things,

1) It moves the reservation clearing stuff into the ordered code, so when
we remove the ordered extent we remove the reservation.
2) It adds a EXTENT_DO_ACCOUNTING flag that gets passed when we clear
delalloc bits in the cases where we want to clear the metadata reservation
when we clear the delalloc extent, in the case that we do an inline extent
or we invalidate the page.
3) It adds another waitqueue to the space info so that when we start a fs
wide delalloc flush, anybody else who also hits that area will simply wait
for the flush to finish and then try to make their allocation.

This has been tested thoroughly to make sure we did not regress on
performance.

Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

1 files changed, 13 insertions, 6 deletions

diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index f9708bd01669..96577e8bf9fd 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -460,7 +460,8 @@ static int clear_state_bit(struct extent_io_tree *tree,
                             struct extent_state *state, int bits, int wake,
                             int delete)
 {
-        int ret = state->state & bits;
+        int bits_to_clear = bits & ~EXTENT_DO_ACCOUNTING;
+        int ret = state->state & bits_to_clear;
 
         if ((bits & EXTENT_DIRTY) && (state->state & EXTENT_DIRTY)) {
                 u64 range = state->end - state->start + 1;
@@ -468,7 +469,7 @@ static int clear_state_bit(struct extent_io_tree *tree,
                 tree->dirty_bytes -= range;
         }
         clear_state_cb(tree, state, bits);
-        state->state &= ~bits;
+        state->state &= ~bits_to_clear;
         if (wake)
                 wake_up(&state->wq);
         if (delete || state->state == 0) {
@@ -956,7 +957,8 @@ int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
                        gfp_t mask)
 {
         return clear_extent_bit(tree, start, end,
-                                EXTENT_DIRTY | EXTENT_DELALLOC, 0, 0,
+                                EXTENT_DIRTY | EXTENT_DELALLOC |
+                                EXTENT_DO_ACCOUNTING, 0, 0,
                                 NULL, mask);
 }
 
@@ -1419,9 +1421,13 @@ int extent_clear_unlock_delalloc(struct inode *inode,
         if (op & EXTENT_CLEAR_DELALLOC)
                 clear_bits |= EXTENT_DELALLOC;
 
+        if (op & EXTENT_CLEAR_ACCOUNTING)
+                clear_bits |= EXTENT_DO_ACCOUNTING;
+
         clear_extent_bit(tree, start, end, clear_bits, 1, 0, NULL, GFP_NOFS);
-        if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK |
-                    EXTENT_END_WRITEBACK | EXTENT_SET_PRIVATE2)))
+        if (!(op & (EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
+                    EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK |
+                    EXTENT_SET_PRIVATE2)))
                 return 0;
 
         while (nr_pages > 0) {
@@ -2709,7 +2715,8 @@ int extent_invalidatepage(struct extent_io_tree *tree,
         lock_extent(tree, start, end, GFP_NOFS);
         wait_on_page_writeback(page);
         clear_extent_bit(tree, start, end,
-                         EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC,
+                         EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
+                         EXTENT_DO_ACCOUNTING,
                          1, 1, NULL, GFP_NOFS);
         return 0;
 }