Re: Review: demand reaping of filestreams objects

[Donald Douwsma <donaldd@xxxxxxx>]

David Chinner wrote:
> Now that the problems with cancel_rearming_delayed_workqueue()
> deadlocks have been fixed, we can go back to only running
> the reaping when we have objects to time out. This prevents
> the reaper from running when there is nothing to do.
> 
> Cheers,
> 
> Dave.

You probably meant to remove the printk's,

@@ -206,8 +206,12 @@ _xfs_mru_cache_list_insert(
...
+               if (!mru->queued) {
+                       printk("reaper started\n");

@@ -271,29 +275,27 @@ _xfs_mru_cache_reap(
+       } else
+       printk("reaper stopped\n");


Apart from that it looks good, much simpler.

Don