Btrfs: nuke fs wide allocation mutex V2

This patch removes the giant fs_info->alloc_mutex and replaces it with a bunch
of little locks.

There is now a pinned_mutex, which is used when messing with the pinned_extents
extent io tree, and the extent_ins_mutex which is used with the pending_del and
extent_ins extent io trees.

The locking for the extent tree stuff was inspired by a patch that Yan Zheng
wrote to fix a race condition, I cleaned it up some and changed the locking
around a little bit, but the idea remains the same.  Basically instead of
holding the extent_ins_mutex throughout the processing of an extent on the
extent_ins or pending_del trees, we just hold it while we're searching and when
we clear the bits on those trees, and lock the extent for the duration of the
operations on the extent.

Also to keep from getting hung up waiting to lock an extent, I've added a
try_lock_extent so if we cannot lock the extent, move on to the next one in the
tree and we'll come back to that one.  I have tested this heavily and it does
not appear to break anything.  This has to be applied on top of my
find_free_extent redo patch.

I tested this patch on top of Yan's space reblancing code and it worked fine.
The only thing that has changed since the last version is I pulled out all my
debugging stuff, apparently I forgot to run guilt refresh before I sent the
last patch out.  Thank you,

Signed-off-by: Josef Bacik <jbacik@redhat.com>

1 files changed, 66 insertions, 26 deletions

diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index 96241f01fa0a..f4926c0f3c8c 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -184,8 +184,8 @@ static int link_free_space(struct btrfs_block_group_cache *block_group,
         return ret;
 }
 
-int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
-                         u64 offset, u64 bytes)
+static int __btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
+                                  u64 offset, u64 bytes)
 {
         struct btrfs_free_space *right_info;
         struct btrfs_free_space *left_info;
@@ -202,8 +202,6 @@ int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
          * are adding, if there is remove that struct and add a new one to
          * cover the entire range
          */
-        spin_lock(&block_group->lock);
-
         right_info = tree_search_offset(&block_group->free_space_offset,
                                         offset+bytes, 0, 1);
         left_info = tree_search_offset(&block_group->free_space_offset,
@@ -261,7 +259,6 @@ int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
         if (ret)
                 kfree(info);
 out:
-        spin_unlock(&block_group->lock);
         if (ret) {
                 printk(KERN_ERR "btrfs: unable to add free space :%d\n", ret);
                 if (ret == -EEXIST)
@@ -274,13 +271,13 @@ out:
         return ret;
 }
 
-int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
-                            u64 offset, u64 bytes)
+static int
+__btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
+                          u64 offset, u64 bytes)
 {
         struct btrfs_free_space *info;
         int ret = 0;
 
-        spin_lock(&block_group->lock);
         info = tree_search_offset(&block_group->free_space_offset, offset, 0,
                                   1);
 
@@ -334,17 +331,63 @@ int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
                 /* step two, insert a new info struct to cover anything
                  * before the hole
                  */
-                spin_unlock(&block_group->lock);
-                ret = btrfs_add_free_space(block_group, old_start,
-                                           offset - old_start);
+                ret = __btrfs_add_free_space(block_group, old_start,
+                                             offset - old_start);
                 BUG_ON(ret);
-                goto out_nolock;
         } else {
                 WARN_ON(1);
         }
 out:
-        spin_unlock(&block_group->lock);
-out_nolock:
+        return ret;
+}
+
+int btrfs_add_free_space(struct btrfs_block_group_cache *block_group,
+                         u64 offset, u64 bytes)
+{
+        int ret;
+        struct btrfs_free_space *sp;
+
+        mutex_lock(&block_group->alloc_mutex);
+        ret = __btrfs_add_free_space(block_group, offset, bytes);
+        sp = tree_search_offset(&block_group->free_space_offset, offset, 0, 1);
+        BUG_ON(!sp);
+        mutex_unlock(&block_group->alloc_mutex);
+
+        return ret;
+}
+
+int btrfs_add_free_space_lock(struct btrfs_block_group_cache *block_group,
+                              u64 offset, u64 bytes)
+{
+        int ret;
+        struct btrfs_free_space *sp;
+
+        ret = __btrfs_add_free_space(block_group, offset, bytes);
+        sp = tree_search_offset(&block_group->free_space_offset, offset, 0, 1);
+        BUG_ON(!sp);
+
+        return ret;
+}
+
+int btrfs_remove_free_space(struct btrfs_block_group_cache *block_group,
+                            u64 offset, u64 bytes)
+{
+        int ret = 0;
+
+        mutex_lock(&block_group->alloc_mutex);
+        ret = __btrfs_remove_free_space(block_group, offset, bytes);
+        mutex_unlock(&block_group->alloc_mutex);
+
+        return ret;
+}
+
+int btrfs_remove_free_space_lock(struct btrfs_block_group_cache *block_group,
+                                 u64 offset, u64 bytes)
+{
+        int ret;
+
+        ret = __btrfs_remove_free_space(block_group, offset, bytes);
+
         return ret;
 }
 
@@ -386,18 +429,18 @@ void btrfs_remove_free_space_cache(struct btrfs_block_group_cache *block_group)
         struct btrfs_free_space *info;
         struct rb_node *node;
 
-        spin_lock(&block_group->lock);
+        mutex_lock(&block_group->alloc_mutex);
         while ((node = rb_last(&block_group->free_space_bytes)) != NULL) {
                 info = rb_entry(node, struct btrfs_free_space, bytes_index);
                 unlink_free_space(block_group, info);
                 kfree(info);
                 if (need_resched()) {
-                        spin_unlock(&block_group->lock);
+                        mutex_unlock(&block_group->alloc_mutex);
                         cond_resched();
-                        spin_lock(&block_group->lock);
+                        mutex_lock(&block_group->alloc_mutex);
                 }
         }
-        spin_unlock(&block_group->lock);
+        mutex_unlock(&block_group->alloc_mutex);
 }
 
 struct btrfs_free_space *btrfs_find_free_space_offset(struct
@@ -407,10 +450,10 @@ struct btrfs_free_space *btrfs_find_free_space_offset(struct
 {
         struct btrfs_free_space *ret;
 
-        spin_lock(&block_group->lock);
+        mutex_lock(&block_group->alloc_mutex);
         ret = tree_search_offset(&block_group->free_space_offset, offset,
                                  bytes, 0);
-        spin_unlock(&block_group->lock);
+        mutex_unlock(&block_group->alloc_mutex);
 
         return ret;
 }
@@ -422,10 +465,10 @@ struct btrfs_free_space *btrfs_find_free_space_bytes(struct
 {
         struct btrfs_free_space *ret;
 
-        spin_lock(&block_group->lock);
+        mutex_lock(&block_group->alloc_mutex);
 
         ret = tree_search_bytes(&block_group->free_space_bytes, offset, bytes);
-        spin_unlock(&block_group->lock);
+        mutex_unlock(&block_group->alloc_mutex);
 
         return ret;
 }
@@ -434,16 +477,13 @@ struct btrfs_free_space *btrfs_find_free_space(struct btrfs_block_group_cache
                                                *block_group, u64 offset,
                                                u64 bytes)
 {
-        struct btrfs_free_space *ret;
+        struct btrfs_free_space *ret = NULL;
 
-        spin_lock(&block_group->lock);
         ret = tree_search_offset(&block_group->free_space_offset, offset,
                                  bytes, 0);
         if (!ret)
                 ret = tree_search_bytes(&block_group->free_space_bytes,
                                         offset, bytes);
 
-        spin_unlock(&block_group->lock);
-
         return ret;
 }