Merge branch 'mpi-master' into wip-k-fmlpwip-k-fmlp

Conflicts:
	litmus/sched_cedf.c

1 files changed, 415 insertions, 391 deletions

diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index d74e6af9b53a..7055d11c1efd 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -10,6 +10,8 @@
 #include <linux/swap.h>
 #include <linux/writeback.h>
 #include <linux/pagevec.h>
+#include <linux/prefetch.h>
+#include <linux/cleancache.h>
 #include "extent_io.h"
 #include "extent_map.h"
 #include "compat.h"
@@ -101,10 +103,10 @@ void extent_io_exit(void)
 }
 
 void extent_io_tree_init(struct extent_io_tree *tree,
-                          struct address_space *mapping, gfp_t mask)
+                         struct address_space *mapping)
 {
         tree->state = RB_ROOT;
-        tree->buffer = RB_ROOT;
+        INIT_RADIX_TREE(&tree->buffer, GFP_ATOMIC);
         tree->ops = NULL;
         tree->dirty_bytes = 0;
         spin_lock_init(&tree->lock);
@@ -235,50 +237,6 @@ static inline struct rb_node *tree_search(struct extent_io_tree *tree,
         return ret;
 }
 
-static struct extent_buffer *buffer_tree_insert(struct extent_io_tree *tree,
-                                          u64 offset, struct rb_node *node)
-{
-        struct rb_root *root = &tree->buffer;
-        struct rb_node **p = &root->rb_node;
-        struct rb_node *parent = NULL;
-        struct extent_buffer *eb;
-
-        while (*p) {
-                parent = *p;
-                eb = rb_entry(parent, struct extent_buffer, rb_node);
-
-                if (offset < eb->start)
-                        p = &(*p)->rb_left;
-                else if (offset > eb->start)
-                        p = &(*p)->rb_right;
-                else
-                        return eb;
-        }
-
-        rb_link_node(node, parent, p);
-        rb_insert_color(node, root);
-        return NULL;
-}
-
-static struct extent_buffer *buffer_search(struct extent_io_tree *tree,
-                                           u64 offset)
-{
-        struct rb_root *root = &tree->buffer;
-        struct rb_node *n = root->rb_node;
-        struct extent_buffer *eb;
-
-        while (n) {
-                eb = rb_entry(n, struct extent_buffer, rb_node);
-                if (offset < eb->start)
-                        n = n->rb_left;
-                else if (offset > eb->start)
-                        n = n->rb_right;
-                else
-                        return eb;
-        }
-        return NULL;
-}
-
 static void merge_cb(struct extent_io_tree *tree, struct extent_state *new,
                      struct extent_state *other)
 {
@@ -483,6 +441,15 @@ static int clear_state_bit(struct extent_io_tree *tree,
         return ret;
 }
 
+static struct extent_state *
+alloc_extent_state_atomic(struct extent_state *prealloc)
+{
+        if (!prealloc)
+                prealloc = alloc_extent_state(GFP_ATOMIC);
+
+        return prealloc;
+}
+
 /*
  * clear some bits on a range in the tree.  This may require splitting
  * or inserting elements in the tree, so the gfp mask is used to
@@ -573,8 +540,8 @@ hit_next:
          */
 
         if (state->start < start) {
-                if (!prealloc)
-                        prealloc = alloc_extent_state(GFP_ATOMIC);
+                prealloc = alloc_extent_state_atomic(prealloc);
+                BUG_ON(!prealloc);
                 err = split_state(tree, state, prealloc, start);
                 BUG_ON(err == -EEXIST);
                 prealloc = NULL;
@@ -595,8 +562,8 @@ hit_next:
          * on the first half
          */
         if (state->start <= end && state->end > end) {
-                if (!prealloc)
-                        prealloc = alloc_extent_state(GFP_ATOMIC);
+                prealloc = alloc_extent_state_atomic(prealloc);
+                BUG_ON(!prealloc);
                 err = split_state(tree, state, prealloc, end + 1);
                 BUG_ON(err == -EEXIST);
                 if (wake)
@@ -734,6 +701,15 @@ static void cache_state(struct extent_state *state,
         }
 }
 
+static void uncache_state(struct extent_state **cached_ptr)
+{
+        if (cached_ptr && (*cached_ptr)) {
+                struct extent_state *state = *cached_ptr;
+                *cached_ptr = NULL;
+                free_extent_state(state);
+        }
+}
+
 /*
  * set some bits on a range in the tree.  This may require allocations or
  * sleeping, so the gfp mask is used to indicate what is allowed.
@@ -760,8 +736,7 @@ int set_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
 again:
         if (!prealloc && (mask & __GFP_WAIT)) {
                 prealloc = alloc_extent_state(mask);
-                if (!prealloc)
-                        return -ENOMEM;
+                BUG_ON(!prealloc);
         }
 
         spin_lock(&tree->lock);
@@ -778,6 +753,8 @@ again:
          */
         node = tree_search(tree, start);
         if (!node) {
+                prealloc = alloc_extent_state_atomic(prealloc);
+                BUG_ON(!prealloc);
                 err = insert_state(tree, prealloc, start, end, &bits);
                 prealloc = NULL;
                 BUG_ON(err == -EEXIST);
@@ -806,20 +783,18 @@ hit_next:
                 if (err)
                         goto out;
 
+                next_node = rb_next(node);
                 cache_state(state, cached_state);
                 merge_state(tree, state);
                 if (last_end == (u64)-1)
                         goto out;
 
                 start = last_end + 1;
-                if (start < end && prealloc && !need_resched()) {
-                        next_node = rb_next(node);
-                        if (next_node) {
-                                state = rb_entry(next_node, struct extent_state,
-                                                 rb_node);
-                                if (state->start == start)
-                                        goto hit_next;
-                        }
+                if (next_node && start < end && prealloc && !need_resched()) {
+                        state = rb_entry(next_node, struct extent_state,
+                                         rb_node);
+                        if (state->start == start)
+                                goto hit_next;
                 }
                 goto search_again;
         }
@@ -846,6 +821,9 @@ hit_next:
                         err = -EEXIST;
                         goto out;
                 }
+
+                prealloc = alloc_extent_state_atomic(prealloc);
+                BUG_ON(!prealloc);
                 err = split_state(tree, state, prealloc, start);
                 BUG_ON(err == -EEXIST);
                 prealloc = NULL;
@@ -876,14 +854,25 @@ hit_next:
                         this_end = end;
                 else
                         this_end = last_start - 1;
+
+                prealloc = alloc_extent_state_atomic(prealloc);
+                BUG_ON(!prealloc);
+
+                /*
+                 * Avoid to free 'prealloc' if it can be merged with
+                 * the later extent.
+                 */
+                atomic_inc(&prealloc->refs);
                 err = insert_state(tree, prealloc, start, this_end,
                                    &bits);
                 BUG_ON(err == -EEXIST);
                 if (err) {
+                        free_extent_state(prealloc);
                         prealloc = NULL;
                         goto out;
                 }
                 cache_state(prealloc, cached_state);
+                free_extent_state(prealloc);
                 prealloc = NULL;
                 start = this_end + 1;
                 goto search_again;
@@ -900,6 +889,9 @@ hit_next:
                         err = -EEXIST;
                         goto out;
                 }
+
+                prealloc = alloc_extent_state_atomic(prealloc);
+                BUG_ON(!prealloc);
                 err = split_state(tree, state, prealloc, end + 1);
                 BUG_ON(err == -EEXIST);
 
@@ -976,18 +968,11 @@ int set_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
                               NULL, mask);
 }
 
-static int clear_extent_new(struct extent_io_tree *tree, u64 start, u64 end,
-                       gfp_t mask)
-{
-        return clear_extent_bit(tree, start, end, EXTENT_NEW, 0, 0,
-                                NULL, mask);
-}
-
 int set_extent_uptodate(struct extent_io_tree *tree, u64 start, u64 end,
-                        gfp_t mask)
+                        struct extent_state **cached_state, gfp_t mask)
 {
-        return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0, NULL,
-                              NULL, mask);
+        return set_extent_bit(tree, start, end, EXTENT_UPTODATE, 0,
+                              NULL, cached_state, mask);
 }
 
 static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
@@ -998,11 +983,6 @@ static int clear_extent_uptodate(struct extent_io_tree *tree, u64 start,
                                 cached_state, mask);
 }
 
-int wait_on_extent_writeback(struct extent_io_tree *tree, u64 start, u64 end)
-{
-        return wait_extent_bit(tree, start, end, EXTENT_WRITEBACK);
-}
-
 /*
  * either insert or lock state struct between start and end use mask to tell
  * us if waiting is desired.
@@ -1056,33 +1036,13 @@ int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end,
                                 mask);
 }
 
-int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end,
-                  gfp_t mask)
+int unlock_extent(struct extent_io_tree *tree, u64 start, u64 end, gfp_t mask)
 {
         return clear_extent_bit(tree, start, end, EXTENT_LOCKED, 1, 0, NULL,
                                 mask);
 }
 
 /*
- * helper function to set pages and extents in the tree dirty
- */
-int set_range_dirty(struct extent_io_tree *tree, u64 start, u64 end)
-{
-        unsigned long index = start >> PAGE_CACHE_SHIFT;
-        unsigned long end_index = end >> PAGE_CACHE_SHIFT;
-        struct page *page;
-
-        while (index <= end_index) {
-                page = find_get_page(tree->mapping, index);
-                BUG_ON(!page);
-                __set_page_dirty_nobuffers(page);
-                page_cache_release(page);
-                index++;
-        }
-        return 0;
-}
-
-/*
  * helper function to set both pages and extents in the tree writeback
  */
 static int set_range_writeback(struct extent_io_tree *tree, u64 start, u64 end)
@@ -1477,12 +1437,13 @@ int extent_clear_unlock_delalloc(struct inode *inode,
  */
 u64 count_range_bits(struct extent_io_tree *tree,
                      u64 *start, u64 search_end, u64 max_bytes,
-                     unsigned long bits)
+                     unsigned long bits, int contig)
 {
         struct rb_node *node;
         struct extent_state *state;
         u64 cur_start = *start;
         u64 total_bytes = 0;
+        u64 last = 0;
         int found = 0;
 
         if (search_end <= cur_start) {
@@ -1507,15 +1468,20 @@ u64 count_range_bits(struct extent_io_tree *tree,
                 state = rb_entry(node, struct extent_state, rb_node);
                 if (state->start > search_end)
                         break;
-                if (state->end >= cur_start && (state->state & bits)) {
+                if (contig && found && state->start > last + 1)
+                        break;
+                if (state->end >= cur_start && (state->state & bits) == bits) {
                         total_bytes += min(search_end, state->end) + 1 -
                                        max(cur_start, state->start);
                         if (total_bytes >= max_bytes)
                                 break;
                         if (!found) {
-                                *start = state->start;
+                                *start = max(cur_start, state->start);
                                 found = 1;
                         }
+                        last = state->end;
+                } else if (contig && found) {
+                        break;
                 }
                 node = rb_next(node);
                 if (!node)
@@ -1773,6 +1739,9 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
 
         do {
                 struct page *page = bvec->bv_page;
+                struct extent_state *cached = NULL;
+                struct extent_state *state;
+
                 tree = &BTRFS_I(page->mapping->host)->io_tree;
 
                 start = ((u64)page->index << PAGE_CACHE_SHIFT) +
@@ -1787,9 +1756,20 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
                 if (++bvec <= bvec_end)
                         prefetchw(&bvec->bv_page->flags);
 
+                spin_lock(&tree->lock);
+                state = find_first_extent_bit_state(tree, start, EXTENT_LOCKED);
+                if (state && state->start == start) {
+                        /*
+                         * take a reference on the state, unlock will drop
+                         * the ref
+                         */
+                        cache_state(state, &cached);
+                }
+                spin_unlock(&tree->lock);
+
                 if (uptodate && tree->ops && tree->ops->readpage_end_io_hook) {
                         ret = tree->ops->readpage_end_io_hook(page, start, end,
-                                                              NULL);
+                                                              state);
                         if (ret)
                                 uptodate = 0;
                 }
@@ -1802,15 +1782,16 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
                                         test_bit(BIO_UPTODATE, &bio->bi_flags);
                                 if (err)
                                         uptodate = 0;
+                                uncache_state(&cached);
                                 continue;
                         }
                 }
 
                 if (uptodate) {
-                        set_extent_uptodate(tree, start, end,
+                        set_extent_uptodate(tree, start, end, &cached,
                                             GFP_ATOMIC);
                 }
-                unlock_extent(tree, start, end, GFP_ATOMIC);
+                unlock_extent_cached(tree, start, end, &cached, GFP_ATOMIC);
 
                 if (whole_page) {
                         if (uptodate) {
@@ -1834,47 +1815,9 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
         bio_put(bio);
 }
 
-/*
- * IO done from prepare_write is pretty simple, we just unlock
- * the structs in the extent tree when done, and set the uptodate bits
- * as appropriate.
- */
-static void end_bio_extent_preparewrite(struct bio *bio, int err)
-{
-        const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-        struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
-        struct extent_io_tree *tree;
-        u64 start;
-        u64 end;
-
-        do {
-                struct page *page = bvec->bv_page;
-                tree = &BTRFS_I(page->mapping->host)->io_tree;
-
-                start = ((u64)page->index << PAGE_CACHE_SHIFT) +
-                        bvec->bv_offset;
-                end = start + bvec->bv_len - 1;
-
-                if (--bvec >= bio->bi_io_vec)
-                        prefetchw(&bvec->bv_page->flags);
-
-                if (uptodate) {
-                        set_extent_uptodate(tree, start, end, GFP_ATOMIC);
-                } else {
-                        ClearPageUptodate(page);
-                        SetPageError(page);
-                }
-
-                unlock_extent(tree, start, end, GFP_ATOMIC);
-
-        } while (bvec >= bio->bi_io_vec);
-
-        bio_put(bio);
-}
-
-static struct bio *
-extent_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
-                 gfp_t gfp_flags)
+struct bio *
+btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
+                gfp_t gfp_flags)
 {
         struct bio *bio;
 
@@ -1901,17 +1844,15 @@ static int submit_one_bio(int rw, struct bio *bio, int mirror_num,
         struct page *page = bvec->bv_page;
         struct extent_io_tree *tree = bio->bi_private;
         u64 start;
-        u64 end;
 
         start = ((u64)page->index << PAGE_CACHE_SHIFT) + bvec->bv_offset;
-        end = start + bvec->bv_len - 1;
 
         bio->bi_private = NULL;
 
         bio_get(bio);
 
         if (tree->ops && tree->ops->submit_bio_hook)
-                tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
+                ret = tree->ops->submit_bio_hook(page->mapping->host, rw, bio,
                                            mirror_num, bio_flags, start);
         else
                 submit_bio(rw, bio);
@@ -1965,7 +1906,9 @@ static int submit_extent_page(int rw, struct extent_io_tree *tree,
         else
                 nr = bio_get_nr_vecs(bdev);
 
-        bio = extent_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
+        bio = btrfs_bio_alloc(bdev, sector, nr, GFP_NOFS | __GFP_HIGH);
+        if (!bio)
+                return -ENOMEM;
 
         bio_add_page(bio, page, page_size, offset);
         bio->bi_end_io = end_io_func;
@@ -1990,6 +1933,7 @@ void set_page_extent_mapped(struct page *page)
 
 static void set_page_extent_head(struct page *page, unsigned long len)
 {
+        WARN_ON(!PagePrivate(page));
         set_page_private(page, EXTENT_PAGE_PRIVATE_FIRST_PAGE | len << 2);
 }
 
@@ -2019,7 +1963,7 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
         struct btrfs_ordered_extent *ordered;
         int ret;
         int nr = 0;
-        size_t page_offset = 0;
+        size_t pg_offset = 0;
         size_t iosize;
         size_t disk_io_size;
         size_t blocksize = inode->i_sb->s_blocksize;
@@ -2027,6 +1971,13 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
 
         set_page_extent_mapped(page);
 
+        if (!PageUptodate(page)) {
+                if (cleancache_get_page(page) == 0) {
+                        BUG_ON(blocksize != PAGE_SIZE);
+                        goto out;
+                }
+        }
+
         end = page_end;
         while (1) {
                 lock_extent(tree, start, end, GFP_NOFS);
@@ -2053,19 +2004,22 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
         while (cur <= end) {
                 if (cur >= last_byte) {
                         char *userpage;
-                        iosize = PAGE_CACHE_SIZE - page_offset;
+                        struct extent_state *cached = NULL;
+
+                        iosize = PAGE_CACHE_SIZE - pg_offset;
                         userpage = kmap_atomic(page, KM_USER0);
-                        memset(userpage + page_offset, 0, iosize);
+                        memset(userpage + pg_offset, 0, iosize);
                         flush_dcache_page(page);
                         kunmap_atomic(userpage, KM_USER0);
                         set_extent_uptodate(tree, cur, cur + iosize - 1,
-                                            GFP_NOFS);
-                        unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+                                            &cached, GFP_NOFS);
+                        unlock_extent_cached(tree, cur, cur + iosize - 1,
+                                             &cached, GFP_NOFS);
                         break;
                 }
-                em = get_extent(inode, page, page_offset, cur,
+                em = get_extent(inode, page, pg_offset, cur,
                                 end - cur + 1, 0);
-                if (IS_ERR(em) || !em) {
+                if (IS_ERR_OR_NULL(em)) {
                         SetPageError(page);
                         unlock_extent(tree, cur, end, GFP_NOFS);
                         break;
@@ -2074,8 +2028,11 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
                 BUG_ON(extent_map_end(em) <= cur);
                 BUG_ON(end < cur);
 
-                if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
+                if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
                         this_bio_flag = EXTENT_BIO_COMPRESSED;
+                        extent_set_compress_type(&this_bio_flag,
+                                                 em->compress_type);
+                }
 
                 iosize = min(extent_map_end(em) - cur, end - cur + 1);
                 cur_end = min(extent_map_end(em) - 1, end);
@@ -2097,16 +2054,19 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
                 /* we've found a hole, just zero and go on */
                 if (block_start == EXTENT_MAP_HOLE) {
                         char *userpage;
+                        struct extent_state *cached = NULL;
+
                         userpage = kmap_atomic(page, KM_USER0);
-                        memset(userpage + page_offset, 0, iosize);
+                        memset(userpage + pg_offset, 0, iosize);
                         flush_dcache_page(page);
                         kunmap_atomic(userpage, KM_USER0);
 
                         set_extent_uptodate(tree, cur, cur + iosize - 1,
-                                            GFP_NOFS);
-                        unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
+                                            &cached, GFP_NOFS);
+                        unlock_extent_cached(tree, cur, cur + iosize - 1,
+                                             &cached, GFP_NOFS);
                         cur = cur + iosize;
-                        page_offset += iosize;
+                        pg_offset += iosize;
                         continue;
                 }
                 /* the get_extent function already copied into the page */
@@ -2115,7 +2075,7 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
                         check_page_uptodate(tree, page);
                         unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
                         cur = cur + iosize;
-                        page_offset += iosize;
+                        pg_offset += iosize;
                         continue;
                 }
                 /* we have an inline extent but it didn't get marked up
@@ -2125,7 +2085,7 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
                         SetPageError(page);
                         unlock_extent(tree, cur, cur + iosize - 1, GFP_NOFS);
                         cur = cur + iosize;
-                        page_offset += iosize;
+                        pg_offset += iosize;
                         continue;
                 }
 
@@ -2138,7 +2098,7 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
                         unsigned long pnr = (last_byte >> PAGE_CACHE_SHIFT) + 1;
                         pnr -= page->index;
                         ret = submit_extent_page(READ, tree, page,
-                                         sector, disk_io_size, page_offset,
+                                         sector, disk_io_size, pg_offset,
                                          bdev, bio, pnr,
                                          end_bio_extent_readpage, mirror_num,
                                          *bio_flags,
@@ -2149,8 +2109,9 @@ static int __extent_read_full_page(struct extent_io_tree *tree,
                 if (ret)
                         SetPageError(page);
                 cur = cur + iosize;
-                page_offset += iosize;
+                pg_offset += iosize;
         }
+out:
         if (!nr) {
                 if (!PageError(page))
                         SetPageUptodate(page);
@@ -2169,7 +2130,7 @@ int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
         ret = __extent_read_full_page(tree, page, get_extent, &bio, 0,
                                       &bio_flags);
         if (bio)
-                submit_one_bio(READ, bio, 0, bio_flags);
+                ret = submit_one_bio(READ, bio, 0, bio_flags);
         return ret;
 }
 
@@ -2204,7 +2165,6 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
         u64 last_byte = i_size_read(inode);
         u64 block_start;
         u64 iosize;
-        u64 unlock_start;
         sector_t sector;
         struct extent_state *cached_state = NULL;
         struct extent_map *em;
@@ -2223,10 +2183,12 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
         unsigned long nr_written = 0;
 
         if (wbc->sync_mode == WB_SYNC_ALL)
-                write_flags = WRITE_SYNC_PLUG;
+                write_flags = WRITE_SYNC;
         else
                 write_flags = WRITE;
 
+        trace___extent_writepage(page, inode, wbc);
+
         WARN_ON(!PageLocked(page));
         pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
         if (page->index > end_index ||
@@ -2329,7 +2291,6 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
                 if (tree->ops && tree->ops->writepage_end_io_hook)
                         tree->ops->writepage_end_io_hook(page, start,
                                                          page_end, NULL, 1);
-                unlock_start = page_end + 1;
                 goto done;
         }
 
@@ -2340,12 +2301,11 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
                         if (tree->ops && tree->ops->writepage_end_io_hook)
                                 tree->ops->writepage_end_io_hook(page, cur,
                                                          page_end, NULL, 1);
-                        unlock_start = page_end + 1;
                         break;
                 }
                 em = epd->get_extent(inode, page, pg_offset, cur,
                                      end - cur + 1, 1);
-                if (IS_ERR(em) || !em) {
+                if (IS_ERR_OR_NULL(em)) {
                         SetPageError(page);
                         break;
                 }
@@ -2387,7 +2347,6 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 
                         cur += iosize;
                         pg_offset += iosize;
-                        unlock_start = cur;
                         continue;
                 }
                 /* leave this out until we have a page_mkwrite call */
@@ -2473,7 +2432,6 @@ static int extent_write_cache_pages(struct extent_io_tree *tree,
         pgoff_t index;
         pgoff_t end;            /* Inclusive */
         int scanned = 0;
-        int range_whole = 0;
 
         pagevec_init(&pvec, 0);
         if (wbc->range_cyclic) {
@@ -2482,8 +2440,6 @@ static int extent_write_cache_pages(struct extent_io_tree *tree,
         } else {
                 index = wbc->range_start >> PAGE_CACHE_SHIFT;
                 end = wbc->range_end >> PAGE_CACHE_SHIFT;
-                if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
-                        range_whole = 1;
                 scanned = 1;
         }
 retry:
@@ -2689,7 +2645,7 @@ int extent_readpages(struct extent_io_tree *tree,
                 prefetchw(&page->flags);
                 list_del(&page->lru);
                 if (!add_to_page_cache_lru(page, mapping,
-                                        page->index, GFP_KERNEL)) {
+                                        page->index, GFP_NOFS)) {
                         __extent_read_full_page(tree, page, get_extent,
                                                 &bio, 0, &bio_flags);
                 }
@@ -2728,123 +2684,6 @@ int extent_invalidatepage(struct extent_io_tree *tree,
 }
 
 /*
- * simple commit_write call, set_range_dirty is used to mark both
- * the pages and the extent records as dirty
- */
-int extent_commit_write(struct extent_io_tree *tree,
-                        struct inode *inode, struct page *page,
-                        unsigned from, unsigned to)
-{
-        loff_t pos = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
-
-        set_page_extent_mapped(page);
-        set_page_dirty(page);
-
-        if (pos > inode->i_size) {
-                i_size_write(inode, pos);
-                mark_inode_dirty(inode);
-        }
-        return 0;
-}
-
-int extent_prepare_write(struct extent_io_tree *tree,
-                         struct inode *inode, struct page *page,
-                         unsigned from, unsigned to, get_extent_t *get_extent)
-{
-        u64 page_start = (u64)page->index << PAGE_CACHE_SHIFT;
-        u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
-        u64 block_start;
-        u64 orig_block_start;
-        u64 block_end;
-        u64 cur_end;
-        struct extent_map *em;
-        unsigned blocksize = 1 << inode->i_blkbits;
-        size_t page_offset = 0;
-        size_t block_off_start;
-        size_t block_off_end;
-        int err = 0;
-        int iocount = 0;
-        int ret = 0;
-        int isnew;
-
-        set_page_extent_mapped(page);
-
-        block_start = (page_start + from) & ~((u64)blocksize - 1);
-        block_end = (page_start + to - 1) | (blocksize - 1);
-        orig_block_start = block_start;
-
-        lock_extent(tree, page_start, page_end, GFP_NOFS);
-        while (block_start <= block_end) {
-                em = get_extent(inode, page, page_offset, block_start,
-                                block_end - block_start + 1, 1);
-                if (IS_ERR(em) || !em)
-                        goto err;
-
-                cur_end = min(block_end, extent_map_end(em) - 1);
-                block_off_start = block_start & (PAGE_CACHE_SIZE - 1);
-                block_off_end = block_off_start + blocksize;
-                isnew = clear_extent_new(tree, block_start, cur_end, GFP_NOFS);
-
-                if (!PageUptodate(page) && isnew &&
-                    (block_off_end > to || block_off_start < from)) {
-                        void *kaddr;
-
-                        kaddr = kmap_atomic(page, KM_USER0);
-                        if (block_off_end > to)
-                                memset(kaddr + to, 0, block_off_end - to);
-                        if (block_off_start < from)
-                                memset(kaddr + block_off_start, 0,
-                                       from - block_off_start);
-                        flush_dcache_page(page);
-                        kunmap_atomic(kaddr, KM_USER0);
-                }
-                if ((em->block_start != EXTENT_MAP_HOLE &&
-                     em->block_start != EXTENT_MAP_INLINE) &&
-                    !isnew && !PageUptodate(page) &&
-                    (block_off_end > to || block_off_start < from) &&
-                    !test_range_bit(tree, block_start, cur_end,
-                                    EXTENT_UPTODATE, 1, NULL)) {
-                        u64 sector;
-                        u64 extent_offset = block_start - em->start;
-                        size_t iosize;
-                        sector = (em->block_start + extent_offset) >> 9;
-                        iosize = (cur_end - block_start + blocksize) &
-                                ~((u64)blocksize - 1);
-                        /*
-                         * we've already got the extent locked, but we
-                         * need to split the state such that our end_bio
-                         * handler can clear the lock.
-                         */
-                        set_extent_bit(tree, block_start,
-                                       block_start + iosize - 1,
-                                       EXTENT_LOCKED, 0, NULL, NULL, GFP_NOFS);
-                        ret = submit_extent_page(READ, tree, page,
-                                         sector, iosize, page_offset, em->bdev,
-                                         NULL, 1,
-                                         end_bio_extent_preparewrite, 0,
-                                         0, 0);
-                        iocount++;
-                        block_start = block_start + iosize;
-                } else {
-                        set_extent_uptodate(tree, block_start, cur_end,
-                                            GFP_NOFS);
-                        unlock_extent(tree, block_start, cur_end, GFP_NOFS);
-                        block_start = cur_end + 1;
-                }
-                page_offset = block_start & (PAGE_CACHE_SIZE - 1);
-                free_extent_map(em);
-        }
-        if (iocount) {
-                wait_extent_bit(tree, orig_block_start,
-                                block_end, EXTENT_LOCKED);
-        }
-        check_page_uptodate(tree, page);
-err:
-        /* FIXME, zero out newly allocated blocks on error */
-        return err;
-}
-
-/*
  * a helper for releasepage, this tests for areas of the page that
  * are locked or under IO and drops the related state bits if it is safe
  * to drop the page.
@@ -2867,9 +2706,17 @@ int try_release_extent_state(struct extent_map_tree *map,
                  * at this point we can safely clear everything except the
                  * locked bit and the nodatasum bit
                  */
-                clear_extent_bit(tree, start, end,
+                ret = clear_extent_bit(tree, start, end,
                                  ~(EXTENT_LOCKED | EXTENT_NODATASUM),
                                  0, 0, NULL, mask);
+
+                /* if clear_extent_bit failed for enomem reasons,
+                 * we can't allow the release to continue.
+                 */
+                if (ret < 0)
+                        ret = 0;
+                else
+                        ret = 1;
         }
         return ret;
 }
@@ -2894,7 +2741,7 @@ int try_release_extent_mapping(struct extent_map_tree *map,
                         len = end - start + 1;
                         write_lock(&map->lock);
                         em = lookup_extent_mapping(map, start, len);
-                        if (!em || IS_ERR(em)) {
+                        if (IS_ERR_OR_NULL(em)) {
                                 write_unlock(&map->lock);
                                 break;
                         }
@@ -2922,76 +2769,169 @@ int try_release_extent_mapping(struct extent_map_tree *map,
         return try_release_extent_state(map, tree, page, mask);
 }
 
-sector_t extent_bmap(struct address_space *mapping, sector_t iblock,
-                get_extent_t *get_extent)
+/*
+ * helper function for fiemap, which doesn't want to see any holes.
+ * This maps until we find something past 'last'
+ */
+static struct extent_map *get_extent_skip_holes(struct inode *inode,
+                                                u64 offset,
+                                                u64 last,
+                                                get_extent_t *get_extent)
 {
-        struct inode *inode = mapping->host;
-        struct extent_state *cached_state = NULL;
-        u64 start = iblock << inode->i_blkbits;
-        sector_t sector = 0;
-        size_t blksize = (1 << inode->i_blkbits);
+        u64 sectorsize = BTRFS_I(inode)->root->sectorsize;
         struct extent_map *em;
+        u64 len;
 
-        lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + blksize - 1,
-                         0, &cached_state, GFP_NOFS);
-        em = get_extent(inode, NULL, 0, start, blksize, 0);
-        unlock_extent_cached(&BTRFS_I(inode)->io_tree, start,
-                             start + blksize - 1, &cached_state, GFP_NOFS);
-        if (!em || IS_ERR(em))
-                return 0;
+        if (offset >= last)
+                return NULL;
 
-        if (em->block_start > EXTENT_MAP_LAST_BYTE)
-                goto out;
+        while(1) {
+                len = last - offset;
+                if (len == 0)
+                        break;
+                len = (len + sectorsize - 1) & ~(sectorsize - 1);
+                em = get_extent(inode, NULL, 0, offset, len, 0);
+                if (IS_ERR_OR_NULL(em))
+                        return em;
 
-        sector = (em->block_start + start - em->start) >> inode->i_blkbits;
-out:
-        free_extent_map(em);
-        return sector;
+                /* if this isn't a hole return it */
+                if (!test_bit(EXTENT_FLAG_VACANCY, &em->flags) &&
+                    em->block_start != EXTENT_MAP_HOLE) {
+                        return em;
+                }
+
+                /* this is a hole, advance to the next extent */
+                offset = extent_map_end(em);
+                free_extent_map(em);
+                if (offset >= last)
+                        break;
+        }
+        return NULL;
 }
 
 int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
                 __u64 start, __u64 len, get_extent_t *get_extent)
 {
-        int ret;
+        int ret = 0;
         u64 off = start;
         u64 max = start + len;
         u32 flags = 0;
+        u32 found_type;
+        u64 last;
+        u64 last_for_get_extent = 0;
         u64 disko = 0;
+        u64 isize = i_size_read(inode);
+        struct btrfs_key found_key;
         struct extent_map *em = NULL;
         struct extent_state *cached_state = NULL;
+        struct btrfs_path *path;
+        struct btrfs_file_extent_item *item;
         int end = 0;
-        u64 em_start = 0, em_len = 0;
+        u64 em_start = 0;
+        u64 em_len = 0;
+        u64 em_end = 0;
         unsigned long emflags;
-        ret = 0;
 
         if (len == 0)
                 return -EINVAL;
 
+        path = btrfs_alloc_path();
+        if (!path)
+                return -ENOMEM;
+        path->leave_spinning = 1;
+
+        /*
+         * lookup the last file extent.  We're not using i_size here
+         * because there might be preallocation past i_size
+         */
+        ret = btrfs_lookup_file_extent(NULL, BTRFS_I(inode)->root,
+                                       path, btrfs_ino(inode), -1, 0);
+        if (ret < 0) {
+                btrfs_free_path(path);
+                return ret;
+        }
+        WARN_ON(!ret);
+        path->slots[0]--;
+        item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+                              struct btrfs_file_extent_item);
+        btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
+        found_type = btrfs_key_type(&found_key);
+
+        /* No extents, but there might be delalloc bits */
+        if (found_key.objectid != btrfs_ino(inode) ||
+            found_type != BTRFS_EXTENT_DATA_KEY) {
+                /* have to trust i_size as the end */
+                last = (u64)-1;
+                last_for_get_extent = isize;
+        } else {
+                /*
+                 * remember the start of the last extent.  There are a
+                 * bunch of different factors that go into the length of the
+                 * extent, so its much less complex to remember where it started
+                 */
+                last = found_key.offset;
+                last_for_get_extent = last + 1;
+        }
+        btrfs_free_path(path);
+
+        /*
+         * we might have some extents allocated but more delalloc past those
+         * extents.  so, we trust isize unless the start of the last extent is
+         * beyond isize
+         */
+        if (last < isize) {
+                last = (u64)-1;
+                last_for_get_extent = isize;
+        }
+
         lock_extent_bits(&BTRFS_I(inode)->io_tree, start, start + len, 0,
                          &cached_state, GFP_NOFS);
-        em = get_extent(inode, NULL, 0, off, max - off, 0);
+
+        em = get_extent_skip_holes(inode, off, last_for_get_extent,
+                                   get_extent);
         if (!em)
                 goto out;
         if (IS_ERR(em)) {
                 ret = PTR_ERR(em);
                 goto out;
         }
+
         while (!end) {
-                off = em->start + em->len;
-                if (off >= max)
-                        end = 1;
+                u64 offset_in_extent;
+
+                /* break if the extent we found is outside the range */
+                if (em->start >= max || extent_map_end(em) < off)
+                        break;
 
-                em_start = em->start;
-                em_len = em->len;
+                /*
+                 * get_extent may return an extent that starts before our
+                 * requested range.  We have to make sure the ranges
+                 * we return to fiemap always move forward and don't
+                 * overlap, so adjust the offsets here
+                 */
+                em_start = max(em->start, off);
 
+                /*
+                 * record the offset from the start of the extent
+                 * for adjusting the disk offset below
+                 */
+                offset_in_extent = em_start - em->start;
+                em_end = extent_map_end(em);
+                em_len = em_end - em_start;
+                emflags = em->flags;
                 disko = 0;
                 flags = 0;
 
+                /*
+                 * bump off for our next call to get_extent
+                 */
+                off = extent_map_end(em);
+                if (off >= max)
+                        end = 1;
+
                 if (em->block_start == EXTENT_MAP_LAST_BYTE) {
                         end = 1;
                         flags |= FIEMAP_EXTENT_LAST;
-                } else if (em->block_start == EXTENT_MAP_HOLE) {
-                        flags |= FIEMAP_EXTENT_UNWRITTEN;
                 } else if (em->block_start == EXTENT_MAP_INLINE) {
                         flags |= (FIEMAP_EXTENT_DATA_INLINE |
                                   FIEMAP_EXTENT_NOT_ALIGNED);
@@ -2999,32 +2939,32 @@ int extent_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
                         flags |= (FIEMAP_EXTENT_DELALLOC |
                                   FIEMAP_EXTENT_UNKNOWN);
                 } else {
-                        disko = em->block_start;
+                        disko = em->block_start + offset_in_extent;
                 }
                 if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags))
                         flags |= FIEMAP_EXTENT_ENCODED;
 
-                emflags = em->flags;
                 free_extent_map(em);
                 em = NULL;
+                if ((em_start >= last) || em_len == (u64)-1 ||
+                   (last == (u64)-1 && isize <= em_end)) {
+                        flags |= FIEMAP_EXTENT_LAST;
+                        end = 1;
+                }
 
-                if (!end) {
-                        em = get_extent(inode, NULL, 0, off, max - off, 0);
-                        if (!em)
-                                goto out;
-                        if (IS_ERR(em)) {
-                                ret = PTR_ERR(em);
-                                goto out;
-                        }
-                        emflags = em->flags;
+                /* now scan forward to see if this is really the last extent. */
+                em = get_extent_skip_holes(inode, off, last_for_get_extent,
+                                           get_extent);
+                if (IS_ERR(em)) {
+                        ret = PTR_ERR(em);
+                        goto out;
                 }
-                if (test_bit(EXTENT_FLAG_VACANCY, &emflags)) {
+                if (!em) {
                         flags |= FIEMAP_EXTENT_LAST;
                         end = 1;
                 }
-
                 ret = fiemap_fill_next_extent(fieinfo, em_start, disko,
-                                        em_len, flags);
+                                              em_len, flags);
                 if (ret)
                         goto out_free;
         }
@@ -3078,6 +3018,8 @@ static struct extent_buffer *__alloc_extent_buffer(struct extent_io_tree *tree,
 #endif
 
         eb = kmem_cache_zalloc(extent_buffer_cache, mask);
+        if (eb == NULL)
+                return NULL;
         eb->start = start;
         eb->len = len;
         spin_lock_init(&eb->lock);
@@ -3104,10 +3046,42 @@ static void __free_extent_buffer(struct extent_buffer *eb)
         kmem_cache_free(extent_buffer_cache, eb);
 }
 
+/*
+ * Helper for releasing extent buffer page.
+ */
+static void btrfs_release_extent_buffer_page(struct extent_buffer *eb,
+                                                unsigned long start_idx)
+{
+        unsigned long index;
+        struct page *page;
+
+        if (!eb->first_page)
+                return;
+
+        index = num_extent_pages(eb->start, eb->len);
+        if (start_idx >= index)
+                return;
+
+        do {
+                index--;
+                page = extent_buffer_page(eb, index);
+                if (page)
+                        page_cache_release(page);
+        } while (index != start_idx);
+}
+
+/*
+ * Helper for releasing the extent buffer.
+ */
+static inline void btrfs_release_extent_buffer(struct extent_buffer *eb)
+{
+        btrfs_release_extent_buffer_page(eb, 0);
+        __free_extent_buffer(eb);
+}
+
 struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
                                           u64 start, unsigned long len,
-                                          struct page *page0,
-                                          gfp_t mask)
+                                          struct page *page0)
 {
         unsigned long num_pages = num_extent_pages(start, len);
         unsigned long i;
@@ -3117,18 +3091,18 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
         struct page *p;
         struct address_space *mapping = tree->mapping;
         int uptodate = 1;
+        int ret;
 
-        spin_lock(&tree->buffer_lock);
-        eb = buffer_search(tree, start);
-        if (eb) {
-                atomic_inc(&eb->refs);
-                spin_unlock(&tree->buffer_lock);
+        rcu_read_lock();
+        eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
+        if (eb && atomic_inc_not_zero(&eb->refs)) {
+                rcu_read_unlock();
                 mark_page_accessed(eb->first_page);
                 return eb;
         }
-        spin_unlock(&tree->buffer_lock);
+        rcu_read_unlock();
 
-        eb = __alloc_extent_buffer(tree, start, len, mask);
+        eb = __alloc_extent_buffer(tree, start, len, GFP_NOFS);
         if (!eb)
                 return NULL;
 
@@ -3145,7 +3119,7 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
                 i = 0;
         }
         for (; i < num_pages; i++, index++) {
-                p = find_or_create_page(mapping, index, mask | __GFP_HIGHMEM);
+                p = find_or_create_page(mapping, index, GFP_NOFS | __GFP_HIGHMEM);
                 if (!p) {
                         WARN_ON(1);
                         goto free_eb;
@@ -3160,50 +3134,77 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
                 }
                 if (!PageUptodate(p))
                         uptodate = 0;
-                unlock_page(p);
+
+                /*
+                 * see below about how we avoid a nasty race with release page
+                 * and why we unlock later
+                 */
+                if (i != 0)
+                        unlock_page(p);
         }
         if (uptodate)
                 set_bit(EXTENT_BUFFER_UPTODATE, &eb->bflags);
 
+        ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
+        if (ret)
+                goto free_eb;
+
         spin_lock(&tree->buffer_lock);
-        exists = buffer_tree_insert(tree, start, &eb->rb_node);
-        if (exists) {
+        ret = radix_tree_insert(&tree->buffer, start >> PAGE_CACHE_SHIFT, eb);
+        if (ret == -EEXIST) {
+                exists = radix_tree_lookup(&tree->buffer,
+                                                start >> PAGE_CACHE_SHIFT);
                 /* add one reference for the caller */
                 atomic_inc(&exists->refs);
                 spin_unlock(&tree->buffer_lock);
+                radix_tree_preload_end();
                 goto free_eb;
         }
         /* add one reference for the tree */
         atomic_inc(&eb->refs);
         spin_unlock(&tree->buffer_lock);
+        radix_tree_preload_end();
+
+        /*
+         * there is a race where release page may have
+         * tried to find this extent buffer in the radix
+         * but failed.  It will tell the VM it is safe to
+         * reclaim the, and it will clear the page private bit.
+         * We must make sure to set the page private bit properly
+         * after the extent buffer is in the radix tree so
+         * it doesn't get lost
+         */
+        set_page_extent_mapped(eb->first_page);
+        set_page_extent_head(eb->first_page, eb->len);
+        if (!page0)
+                unlock_page(eb->first_page);
         return eb;
 
 free_eb:
+        if (eb->first_page && !page0)
+                unlock_page(eb->first_page);
+
         if (!atomic_dec_and_test(&eb->refs))
                 return exists;
-        for (index = 1; index < i; index++)
-                page_cache_release(extent_buffer_page(eb, index));
-        page_cache_release(extent_buffer_page(eb, 0));
-        __free_extent_buffer(eb);
+        btrfs_release_extent_buffer(eb);
         return exists;
 }
 
 struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree,
-                                         u64 start, unsigned long len,
-                                          gfp_t mask)
+                                         u64 start, unsigned long len)
 {
         struct extent_buffer *eb;
 
-        spin_lock(&tree->buffer_lock);
-        eb = buffer_search(tree, start);
-        if (eb)
-                atomic_inc(&eb->refs);
-        spin_unlock(&tree->buffer_lock);
-
-        if (eb)
+        rcu_read_lock();
+        eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
+        if (eb && atomic_inc_not_zero(&eb->refs)) {
+                rcu_read_unlock();
                 mark_page_accessed(eb->first_page);
+                return eb;
+        }
+        rcu_read_unlock();
 
-        return eb;
+        return NULL;
 }
 
 void free_extent_buffer(struct extent_buffer *eb)
@@ -3232,10 +3233,11 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree,
                         continue;
 
                 lock_page(page);
+                WARN_ON(!PagePrivate(page));
+
+                set_page_extent_mapped(page);
                 if (i == 0)
                         set_page_extent_head(page, eb->len);
-                else
-                        set_page_private(page, EXTENT_PAGE_PRIVATE);
 
                 clear_page_dirty_for_io(page);
                 spin_lock_irq(&page->mapping->tree_lock);
@@ -3250,13 +3252,6 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree,
         return 0;
 }
 
-int wait_on_extent_buffer_writeback(struct extent_io_tree *tree,
-                                    struct extent_buffer *eb)
-{
-        return wait_on_extent_writeback(tree, eb->start,
-                                        eb->start + eb->len - 1);
-}
-
 int set_extent_buffer_dirty(struct extent_io_tree *tree,
                              struct extent_buffer *eb)
 {
@@ -3302,7 +3297,7 @@ int set_extent_buffer_uptodate(struct extent_io_tree *tree,
         num_pages = num_extent_pages(eb->start, eb->len);
 
         set_extent_uptodate(tree, eb->start, eb->start + eb->len - 1,
-                            GFP_NOFS);
+                            NULL, GFP_NOFS);
         for (i = 0; i < num_pages; i++) {
                 page = extent_buffer_page(eb, i);
                 if ((i == 0 && (eb->start & (PAGE_CACHE_SIZE - 1))) ||
@@ -3425,6 +3420,13 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 
         for (i = start_i; i < num_pages; i++) {
                 page = extent_buffer_page(eb, i);
+
+                WARN_ON(!PagePrivate(page));
+
+                set_page_extent_mapped(page);
+                if (i == 0)
+                        set_page_extent_head(page, eb->len);
+
                 if (inc_all_pages)
                         page_cache_get(page);
                 if (!PageUptodate(page)) {
@@ -3530,6 +3532,7 @@ int map_private_extent_buffer(struct extent_buffer *eb, unsigned long start,
                        "wanted %lu %lu\n", (unsigned long long)eb->start,
                        eb->len, start, min_len);
                 WARN_ON(1);
+                return -EINVAL;
         }
 
         p = extent_buffer_page(eb, i);
@@ -3722,6 +3725,12 @@ static void move_pages(struct page *dst_page, struct page *src_page,
         kunmap_atomic(dst_kaddr, KM_USER0);
 }
 
+static inline bool areas_overlap(unsigned long src, unsigned long dst, unsigned long len)
+{
+        unsigned long distance = (src > dst) ? src - dst : dst - src;
+        return distance < len;
+}
+
 static void copy_pages(struct page *dst_page, struct page *src_page,
                        unsigned long dst_off, unsigned long src_off,
                        unsigned long len)
@@ -3729,10 +3738,12 @@ static void copy_pages(struct page *dst_page, struct page *src_page,
         char *dst_kaddr = kmap_atomic(dst_page, KM_USER0);
         char *src_kaddr;
 
-        if (dst_page != src_page)
+        if (dst_page != src_page) {
                 src_kaddr = kmap_atomic(src_page, KM_USER1);
-        else
+        } else {
                 src_kaddr = dst_kaddr;
+                BUG_ON(areas_overlap(src_off, dst_off, len));
+        }
 
         memcpy(dst_kaddr + dst_off, src_kaddr + src_off, len);
         kunmap_atomic(dst_kaddr, KM_USER0);
@@ -3807,7 +3818,7 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
                        "len %lu len %lu\n", dst_offset, len, dst->len);
                 BUG_ON(1);
         }
-        if (dst_offset < src_offset) {
+        if (!areas_overlap(src_offset, dst_offset, len)) {
                 memcpy_extent_buffer(dst, dst_offset, src_offset, len);
                 return;
         }
@@ -3833,34 +3844,47 @@ void memmove_extent_buffer(struct extent_buffer *dst, unsigned long dst_offset,
         }
 }
 
+static inline void btrfs_release_extent_buffer_rcu(struct rcu_head *head)
+{
+        struct extent_buffer *eb =
+                        container_of(head, struct extent_buffer, rcu_head);
+
+        btrfs_release_extent_buffer(eb);
+}
+
 int try_release_extent_buffer(struct extent_io_tree *tree, struct page *page)
 {
         u64 start = page_offset(page);
         struct extent_buffer *eb;
         int ret = 1;
-        unsigned long i;
-        unsigned long num_pages;
 
         spin_lock(&tree->buffer_lock);
-        eb = buffer_search(tree, start);
-        if (!eb)
-                goto out;
+        eb = radix_tree_lookup(&tree->buffer, start >> PAGE_CACHE_SHIFT);
+        if (!eb) {
+                spin_unlock(&tree->buffer_lock);
+                return ret;
+        }
 
-        if (atomic_read(&eb->refs) > 1) {
+        if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
                 ret = 0;
                 goto out;
         }
-        if (test_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
+
+        /*
+         * set @eb->refs to 0 if it is already 1, and then release the @eb.
+         * Or go back.
+         */
+        if (atomic_cmpxchg(&eb->refs, 1, 0) != 1) {
                 ret = 0;
                 goto out;
         }
-        /* at this point we can safely release the extent buffer */
-        num_pages = num_extent_pages(eb->start, eb->len);
-        for (i = 0; i < num_pages; i++)
-                page_cache_release(extent_buffer_page(eb, i));
-        rb_erase(&eb->rb_node, &tree->buffer);
-        __free_extent_buffer(eb);
+
+        radix_tree_delete(&tree->buffer, start >> PAGE_CACHE_SHIFT);
 out:
         spin_unlock(&tree->buffer_lock);
+
+        /* at this point we can safely release the extent buffer */
+        if (atomic_read(&eb->refs) == 0)
+                call_rcu(&eb->rcu_head, btrfs_release_extent_buffer_rcu);
         return ret;
 }