Btrfs: don't use an async starter for most of our workers
authorJosef Bacik <jbacik@fusionio.com>
Thu, 29 Aug 2013 20:51:22 +0000 (16:51 -0400)
committerChris Mason <chris.mason@fusionio.com>
Sun, 1 Sep 2013 12:16:41 +0000 (08:16 -0400)
We only need an async starter if we can't make a GFP_NOFS allocation in our
current path.  This is the case for the endio stuff since it happens in IRQ
context, but things like the caching thread workers and the delalloc flushers we
can easily make this allocation and start threads right away.  Also change the
worker count for the caching thread pool.  Traditionally we limited this to 2
since we took read locks while caching, but nowadays we do this lockless so
there's no reason to limit the number of caching threads.  Thanks,

Signed-off-by: Josef Bacik <jbacik@fusionio.com>
Signed-off-by: Chris Mason <chris.mason@fusionio.com>
fs/btrfs/disk-io.c

index 21d1d066758ea94ac2ffc7a4d3dff0fe88b128cc..4cbb00af92ff3bed86561b9f02ec6fbcc04665b2 100644 (file)
@@ -2483,20 +2483,17 @@ int open_ctree(struct super_block *sb,
                           &fs_info->generic_worker);
 
        btrfs_init_workers(&fs_info->delalloc_workers, "delalloc",
-                          fs_info->thread_pool_size,
-                          &fs_info->generic_worker);
+                          fs_info->thread_pool_size, NULL);
 
        btrfs_init_workers(&fs_info->flush_workers, "flush_delalloc",
-                          fs_info->thread_pool_size,
-                          &fs_info->generic_worker);
+                          fs_info->thread_pool_size, NULL);
 
        btrfs_init_workers(&fs_info->submit_workers, "submit",
                           min_t(u64, fs_devices->num_devices,
-                          fs_info->thread_pool_size),
-                          &fs_info->generic_worker);
+                          fs_info->thread_pool_size), NULL);
 
        btrfs_init_workers(&fs_info->caching_workers, "cache",
-                          2, &fs_info->generic_worker);
+                          fs_info->thread_pool_size, NULL);
 
        /* a higher idle thresh on the submit workers makes it much more
         * likely that bios will be send down in a sane order to the