Re: [PATCH 3/5] blktrace: refcount the request_queue during ioctl

[Ming Lei <ming.lei@redhat.com>]

On Tue, Apr 14, 2020 at 04:19:00AM +0000, Luis Chamberlain wrote:
> Ensure that the request_queue is refcounted during its full
> ioctl cycle. This avoids possible races against removal, given
> blk_get_queue() also checks to ensure the queue is not dying.
> 
> This small race is possible if you defer removal of the request_queue
> and userspace fires off an ioctl for the device in the meantime.
> 
> Cc: Bart Van Assche <bvanassche@acm.org>
> Cc: Omar Sandoval <osandov@fb.com>
> Cc: Hannes Reinecke <hare@suse.com>
> Cc: Nicolai Stange <nstange@suse.de>
> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Cc: Michal Hocko <mhocko@kernel.org>
> Cc: yu kuai <yukuai3@huawei.com>
> Reviewed-by: Bart Van Assche <bvanassche@acm.org>
> Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
> ---
>  kernel/trace/blktrace.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/kernel/trace/blktrace.c b/kernel/trace/blktrace.c
> index 15086227592f..17e144d15779 100644
> --- a/kernel/trace/blktrace.c
> +++ b/kernel/trace/blktrace.c
> @@ -701,6 +701,9 @@ int blk_trace_ioctl(struct block_device *bdev, unsigned cmd, char __user *arg)
>  	if (!q)
>  		return -ENXIO;
>  
> +	if (!blk_get_queue(q))
> +		return -ENXIO;
> +
>  	mutex_lock(&q->blk_trace_mutex);
>  
>  	switch (cmd) {
> @@ -729,6 +732,9 @@ int blk_trace_ioctl(struct block_device *bdev, unsigned cmd, char __user *arg)
>  	}
>  
>  	mutex_unlock(&q->blk_trace_mutex);
> +
> +	blk_put_queue(q);
> +
>  	return ret;
>  }

Actually when bdev is opened, one extra refcount is held on gendisk, so
gendisk won't go away. And __device_add_disk() does grab one extra
refcount on request queue, so request queue shouldn't go away when ioctl
is running.

Can you describe a bit more what the issue is to be addressed by this
patch?

Thanks,
Ming