Re: [RFC PATCH 1/8] rqbased-dm: allow blk_get_request() to be called from interrupt context

[Jens Axboe]

On Tue, Dec 19 2006, Kiyoshi Ueda wrote:
> Currently, blk_get_request() is not ready for being called from
> interrupt context because it enables interrupt forcibly in it.
> 
> Request-based device-mapper sometimes needs to get a request
> in interrupt context to create a clone.
> Calling blk_get_request() from interrupt context should be OK
> because blk_get_request() returns NULL without sleep if no memory
> unless __GFP_WAIT mask is specified, and then the interrupt context
> can plug queue to retry after and return immediately.
> 
> So this patch allows blk_get_request() to be called from interrupt
> context by save/restore current state of irq.
> 
> 
> Signed-off-by: Kiyoshi Ueda <k-ueda@xxxxxxxxxxxxx>
> Signed-off-by: Jun'ichi Nomura <j-nomura@xxxxxxxxxxxxx>
> 
> diff -rupN 2.6.19.1/block/ll_rw_blk.c 1-blk-get-request-irqrestore/block/ll_rw_blk.c
> --- 2.6.19.1/block/ll_rw_blk.c	2006-12-11 14:32:53.000000000 -0500
> +++ 1-blk-get-request-irqrestore/block/ll_rw_blk.c	2006-12-15 10:21:29.000000000 -0500
> @@ -2064,9 +2064,10 @@ static void freed_request(request_queue_
>   * Get a free request, queue_lock must be held.
>   * Returns NULL on failure, with queue_lock held.
>   * Returns !NULL on success, with queue_lock *not held*.
> + * If flags is given, the irq state is kept when unlocking the queue_lock.
>   */
>  static struct request *get_request(request_queue_t *q, int rw, struct bio *bio,
> -				   gfp_t gfp_mask)
> +				   gfp_t gfp_mask, unsigned long *flags)
>  {
>  	struct request *rq = NULL;
>  	struct request_list *rl = &q->rq;
> @@ -2119,7 +2120,10 @@ static struct request *get_request(reque
>  	if (priv)
>  		rl->elvpriv++;
>  
> -	spin_unlock_irq(q->queue_lock);
> +	if (flags)
> +		spin_unlock_irqrestore(q->queue_lock, *flags);
> +	else
> +		spin_unlock_irq(q->queue_lock);

Big NACK on this - it's not only really ugly, it's also buggy to pass
interrupt flags as function arguments. As you also mention in the 0/1
mail, this also breaks CFQ.

Why do you need in-interrupt request allocation?

-- 
Jens Axboe

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/