Messages in this thread | | | Date | Wed, 24 Jul 2002 15:08:57 +0200 | From | Marcin Dalecki <> | Subject | Re: please DON'T run 2.5.27 with IDE! |
| |
Jens Axboe wrote:
>>>>2.5.27:drivers/block/ll_rw_blk.c >>>>void blk_start_queue(request_queue_t *q) >>>>{ >>>> if (test_and_clear_bit(QUEUE_FLAG_STOPPED, &q->queue_flags)) { >>>> unsigned long flags; >>>> >>>> spin_lock_irqsave(q->queue_lock, flags); >>>> if (!elv_queue_empty(q)) >>>> q->request_fn(q); >>>> spin_unlock_irqrestore(q->queue_lock, flags); >>>> } >>>>}
> There were buggy versions at one point, however they may not have made it > into a full release. In that case it was just bk version of 2.5.19-pre > effectively. I forget the details :-)
Naj - it's far more trivial I just looked at wrong tree at hand... But anyway. What happens if somone does set QUEUE_FLAG_STOPPED between the test_and_claer_bit and taking the spin_lock? Setting the QUEUE_FLAG_STOPPED isn't maintaining the spin_lock protection!
My goal is to make sure that the QUEUE_FLAG_STOPPED has a valid value *inside* the q->request_fn call.
This here is where it's supposed to be set:
void blk_stop_queue(request_queue_t *q) { unsigned long flags;
spin_lock_irqsave(q->queue_lock, flags); blk_remove_plug(q); spin_unlock_irqrestore(q->queue_lock, flags);
set_bit(QUEUE_FLAG_STOPPED, &q->queue_flags); }
And I don't see anything preventing the above problem.
It sould perhaps be?
void blk_stop_queue(request_queue_t *q) { unsigned long flags;
spin_lock_irqsave(q->queue_lock, flags); blk_remove_plug(q); set_bit(QUEUE_FLAG_STOPPED, &q->queue_flags); /* Notice spinlock still held! */ spin_unlock_irqrestore(q->queue_lock, flags); }
void blk_start_queue(request_queue_t *q) { if (test_bit(QUEUE_FLAG_STOPPED, &q->queue_flags)) { unsigned long flags;
spin_lock_irqsave(q->queue_lock, flags); if (!test_bit(QUEUE_FLAG_STOPPED, &q->queue_flags)) { spin_unlock_irqsave(q->queue_lock, flags); return; } clear_bit(QUEUE_FLAG_STOPPED, &q->queue_flags); if (!elv_queue_empty(q)) q->request_fn(q); spin_unlock_irqrestore(q->queue_lock, flags); } }
At least I couldn't see any harm in doing it like above. And again. I think it would assert that the flag is well defined inside q->request_fn().
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |