2010-11-30 17:01:28

by Jerome Marchand

[permalink] [raw]
Subject: Re: [GIT PULL] Revert of the IO stat fix

On 10/24/2010 10:44 PM, Jens Axboe wrote:
> diff --git a/block/genhd.c b/block/genhd.c
> index a8adf96..7d4d860 100644
> --- a/block/genhd.c
> +++ b/block/genhd.c
> @@ -930,14 +930,9 @@ static void disk_free_ptbl_rcu_cb(struct rcu_head *head)
> struct disk_part_tbl *ptbl =
> container_of(head, struct disk_part_tbl, rcu_head);
> struct gendisk *disk = ptbl->disk;
> - struct request_queue *q = disk->queue;
> - unsigned long flags;
>
> kfree(ptbl);
> -
> - spin_lock_irqsave(q->queue_lock, flags);
> - elv_quiesce_end(q);
> - spin_unlock_irqrestore(q->queue_lock, flags);
> + elv_quiesce_end(disk->queue);

Here, the queue may be already gone. We should check it is still here and alive:

static void disk_free_ptbl_rcu_cb(struct rcu_head *head)
{
struct disk_part_tbl *ptbl =
container_of(head, struct disk_part_tbl, rcu_head);
struct gendisk *disk = ptbl->disk;
struct request_queue *q = disk->queue;

kfree(ptbl);
if (q && !test_bit(QUEUE_FLAG_DEAD, &q->queue_flags))
elv_quiesce_end(q);
}


> }
>
> /**
> @@ -962,10 +957,7 @@ static void disk_replace_part_tbl(struct gendisk *disk,
> if (old_ptbl) {
> rcu_assign_pointer(old_ptbl->last_lookup, NULL);
>
> - spin_lock_irq(q->queue_lock);
> elv_quiesce_start(q);

Same as above.

I'm not sure that is enough, but on my test machine these changes fix the crash
at USB key device removal.

Regards,
Jerome

> - spin_unlock_irq(q->queue_lock);
> -
> call_rcu(&old_ptbl->rcu_head, disk_free_ptbl_rcu_cb);
> }
> }