Re: Deadlocks due to per-process plugging

[Thomas Gleixner <tglx@linutronix.de>]

On Mon, 16 Jul 2012, Mike Galbraith wrote:

> Hm, wonder how bad this sucks.. and if I should go hide under a big
> sturdy rock after I poke xmit :)
> 
> ---
>  block/blk-core.c |    1 +
>  kernel/rtmutex.c |   11 +++++++++--
>  2 files changed, 10 insertions(+), 2 deletions(-)
> 
> --- a/block/blk-core.c
> +++ b/block/blk-core.c
> @@ -2782,6 +2782,7 @@ void blk_flush_plug_list(struct blk_plug
>  	if (q)
>  		queue_unplugged(q, depth, from_schedule);
>  }
> +EXPORT_SYMBOL(blk_flush_plug_list);

You don't need that one. blk-core and rtmutex are builtin

>  void blk_finish_plug(struct blk_plug *plug)
>  {
> --- a/kernel/rtmutex.c
> +++ b/kernel/rtmutex.c
> @@ -20,6 +20,7 @@
>  #include <linux/module.h>
>  #include <linux/sched.h>
>  #include <linux/timer.h>
> +#include <linux/blkdev.h>
>  
>  #include "rtmutex_common.h"
>  
> @@ -647,8 +648,11 @@ static inline void rt_spin_lock_fastlock
>  
>  	if (likely(rt_mutex_cmpxchg(lock, NULL, current)))
>  		rt_mutex_deadlock_account_lock(lock, current);
> -	else
> +	else {
> +		if (blk_needs_flush_plug(current))
> +			blk_schedule_flush_plug(current);
>  		slowfn(lock);
> +	}

That should do the trick.