Re: [PATCH 3/4] make nbd device wait for its users.

[Markus Pargmann <mpa@pengutronix.de>]

[-- Attachment #1: Type: text/plain, Size: 4815 bytes --]

Hi,

sorry I couldn't fit the review into last week.

On Tuesday 24 May 2016 14:26:27 Pranay Kr. Srivastava wrote:
> When a timeout occurs or a recv fails, then
> instead of abruplty killing nbd block device
> wait for it's users to finish.
> 
> This is more required when filesystem(s) like
> ext2 or ext3 don't expect their buffer heads to
> disappear while the filesystem is mounted.
> 
> Use a kref for users using this. The device will
> be released for kref count of 2, not less or more.
> 
> Signed-off-by: Pranay Kr. Srivastava <pranjas@gmail.com>
> ---
>  drivers/block/nbd.c |   51 +++++++++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 51 insertions(+)
> 
> diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
> index af86c9b..59db890 100644
> --- a/drivers/block/nbd.c
> +++ b/drivers/block/nbd.c
> @@ -71,6 +71,8 @@ struct nbd_device {
>  	struct dentry *dbg_dir;
>  #endif
>  	struct work_struct ws_nbd;
> +	struct kref users;
> +	struct completion user_completion;
>  };
>  
>  #if IS_ENABLED(CONFIG_DEBUG_FS)
> @@ -674,6 +676,7 @@ static void nbd_reset(struct nbd_device *nbd)
>  	nbd->flags = 0;
>  	nbd->xmit_timeout = 0;
>  	INIT_WORK(&nbd->ws_nbd, nbd_work_func);
> +	init_completion(&nbd->user_completion);
>  	queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, nbd->disk->queue);
>  	del_timer_sync(&nbd->timeout_timer);
>  }
> @@ -807,6 +810,7 @@ static int __nbd_ioctl(struct block_device *bdev, struct nbd_device *nbd,
>  		kthread_stop(thread);
>  
>  		sock_shutdown(nbd);
> +		wait_for_completion(&nbd->user_completion);
>  		mutex_lock(&nbd->tx_lock);
>  		nbd_clear_que(nbd);
>  		kill_bdev(bdev);
> @@ -858,12 +862,58 @@ static int nbd_ioctl(struct block_device *bdev, fmode_t mode,
>  	return error;
>  }
>  
> +static void nbd_kref_release(struct kref *kref_users)
> +{
> +	struct nbd_device *nbd = container_of(kref_users, struct nbd_device,
> +			users);
> +	pr_debug("Releasing kref [%s]\n", __FUNCTION__);
> +	complete(&nbd->user_completion);
> +
> +}
> +
> +static int nbd_open(struct block_device *bdev, fmode_t mode)
> +{
> +	struct nbd_device *nbd_dev = bdev->bd_disk->private_data;
> +
> +	kref_get(&nbd_dev->users);
> +	pr_debug("Opening nbd_dev %s. Active users = %u\n",
> +			bdev->bd_disk->disk_name,
> +			atomic_read(&nbd_dev->users.refcount) - 1);
> +	return 0;
> +}
> +
> +static void nbd_release(struct gendisk *disk, fmode_t mode)
> +{
> +	struct nbd_device *nbd_dev = disk->private_data;
> +	/*
> +	*kref_init initializes ref count to 1, so we
> +	*we check for refcount to be 2 for a final put.
> +	*
> +	*kref needs to be re-initialized just here as the
> +	*other process holding it must see the ref count as 2.
> +	*/
> +	kref_put(&nbd_dev->users,  nbd_kref_release);
> +
> +	if (atomic_read(&nbd_dev->users.refcount) == 2) {
> +		kref_sub(&nbd_dev->users, 2, nbd_kref_release);
> +		kref_init(&nbd_dev->users);
> +		kref_get(&nbd_dev->users);

Reading the refcount directly seems not to be as it supposed to be.

Why don't you put a kref_init() and kref_put() call into NBD_DO_IT? This
way you don't have to work around the property that kref_init() starts
with a refcount of 1 but you can use it.

For example:
	NBD_DO_IT:
		kref_init()
		...
		kref_put()

	nbd_thread_recv() and nbd_thread_send():
		kref_get()
		...
		kref_put()

	In nbd_open() you could use kref_get_unless_zero() to avoid
	opening a not connected device.

	nbd_release() would then be a very simple kref_put() without
	checking for 2 and so on.

Also there are some checkpatch issues with this patch.

Best Regards,

Markus

> +	}
> +
> +	pr_debug("Closing nbd_dev %s. Active users = %u\n",
> +			disk->disk_name,
> +			atomic_read(&nbd_dev->users.refcount) - 1);
> +}
> +
>  static const struct block_device_operations nbd_fops = {
>  	.owner =	THIS_MODULE,
>  	.ioctl =	nbd_ioctl,
>  	.compat_ioctl =	nbd_ioctl,
> +	.open = 	nbd_open,
> +	.release = 	nbd_release
>  };
>  
> +
>  static void nbd_work_func(struct work_struct *ws_nbd)
>  {
>  	struct nbd_device *nbd_dev = container_of(ws_nbd, struct nbd_device,
> @@ -1098,6 +1148,7 @@ static int __init nbd_init(void)
>  		disk->first_minor = i << part_shift;
>  		disk->fops = &nbd_fops;
>  		disk->private_data = &nbd_dev[i];
> +		kref_init(&nbd_dev[i].users);
>  		sprintf(disk->disk_name, "nbd%d", i);
>  		nbd_reset(&nbd_dev[i]);
>  		add_disk(disk);
> 

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 819 bytes --]