patch (Re: 3.18.3 + latest bcache-dev)

["Stephen R. van den Berg" <srb@cuci.nl>]

Stephen R. van den Berg wrote:
>What I notice are:
>- The 100% CPU usage for four bcache_writebac processes is back.  I've
>  already tried raising the value you suggested from 30 to 60, but I did
>  not have an opportunity to reboot yet.  Will report back on this.

What I saw:
a. After running several hours, copying roughly 3.0TB through the two
   SSD caching devices onto the BTRFS backing device consisting of
   four HDDs, the bcache_writebac processes were still eating 100% CPU each.
b. Then did a manual umount of the fs, then stopped the bcache device
   through the proc filesystem.
c. Rebooted.
d. Came back in a patched kernel, 100% CPU usage is gone.

The patch I used (I did not investigate the correctness of the value
picked, I simply doubled it as an initial guess):

commit 2abecdbd214c94741012de32759bb0de4c6adbf9
Author: Stephen R. van den Berg <srb@cuci.nl>
Date:   Tue Jan 13 23:24:52 2015 +0100

    Increase btree_scan_ratelimit to avoid 100% CPU hogging.

diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
index 653419b..5fff8ad 100644
--- a/drivers/md/bcache/super.c
+++ b/drivers/md/bcache/super.c
@@ -867,7 +867,7 @@ static struct cache_set *bch_cache_set_alloc(struct cache *ca)
 	c->error_limit	= 16 << IO_ERROR_SHIFT;
 
 	c->tiering_percent = 10;
-	c->btree_scan_ratelimit = 30;
+	c->btree_scan_ratelimit = 60;
 
 	c->copy_gc_enabled = 1;
 	c->tiering_enabled = 1;
-- 
Stephen.