Re: [PATCH 5.9 374/391] null_blk: synchronization fix for zoned device

[Damien Le Moal]

On 2020/11/04 5:52, Greg Kroah-Hartman wrote:
> From: Kanchan Joshi <joshi.k@xxxxxxxxxxx>
> 
> commit 35bc10b2eafbb701064b94f283b77c54d3304842 upstream.
> 
> Parallel write,read,zone-mgmt operations accessing/altering zone state
> and write-pointer may get into race. Avoid the situation by using a new
> spinlock for zoned device.
> Concurrent zone-appends (on a zone) returning same write-pointer issue
> is also avoided using this lock.
> 
> Cc: stable@xxxxxxxxxxxxxxx
> Fixes: e0489ed5daeb ("null_blk: Support REQ_OP_ZONE_APPEND")
> Signed-off-by: Kanchan Joshi <joshi.k@xxxxxxxxxxx>
> Reviewed-by: Damien Le Moal <damien.lemoal@xxxxxxx>
> Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>

Greg,

I sent a followup patch fixing a bug introduced by this patch, but I forgot to
mark it for stable. The patch is

commit aa1c09cb65e2 "null_blk: Fix locking in zoned mode"

Could you pickup that one too please ?

Best regards.

>
> ---
>  drivers/block/null_blk.h       |    1 +
>  drivers/block/null_blk_zoned.c |   22 ++++++++++++++++++----
>  2 files changed, 19 insertions(+), 4 deletions(-)
> 
> --- a/drivers/block/null_blk.h
> +++ b/drivers/block/null_blk.h
> @@ -44,6 +44,7 @@ struct nullb_device {
>  	unsigned int nr_zones;
>  	struct blk_zone *zones;
>  	sector_t zone_size_sects;
> +	spinlock_t zone_lock;
>  
>  	unsigned long size; /* device size in MB */
>  	unsigned long completion_nsec; /* time in ns to complete a request */
> --- a/drivers/block/null_blk_zoned.c
> +++ b/drivers/block/null_blk_zoned.c
> @@ -45,6 +45,7 @@ int null_init_zoned_dev(struct nullb_dev
>  	if (!dev->zones)
>  		return -ENOMEM;
>  
> +	spin_lock_init(&dev->zone_lock);
>  	if (dev->zone_nr_conv >= dev->nr_zones) {
>  		dev->zone_nr_conv = dev->nr_zones - 1;
>  		pr_info("changed the number of conventional zones to %u",
> @@ -131,8 +132,11 @@ int null_report_zones(struct gendisk *di
>  		 * So use a local copy to avoid corruption of the device zone
>  		 * array.
>  		 */
> +		spin_lock_irq(&dev->zone_lock);
>  		memcpy(&zone, &dev->zones[first_zone + i],
>  		       sizeof(struct blk_zone));
> +		spin_unlock_irq(&dev->zone_lock);
> +
>  		error = cb(&zone, i, data);
>  		if (error)
>  			return error;
> @@ -277,18 +281,28 @@ static blk_status_t null_zone_mgmt(struc
>  blk_status_t null_process_zoned_cmd(struct nullb_cmd *cmd, enum req_opf op,
>  				    sector_t sector, sector_t nr_sectors)
>  {
> +	blk_status_t sts;
> +	struct nullb_device *dev = cmd->nq->dev;
> +
> +	spin_lock_irq(&dev->zone_lock);
>  	switch (op) {
>  	case REQ_OP_WRITE:
> -		return null_zone_write(cmd, sector, nr_sectors, false);
> +		sts = null_zone_write(cmd, sector, nr_sectors, false);
> +		break;
>  	case REQ_OP_ZONE_APPEND:
> -		return null_zone_write(cmd, sector, nr_sectors, true);
> +		sts = null_zone_write(cmd, sector, nr_sectors, true);
> +		break;
>  	case REQ_OP_ZONE_RESET:
>  	case REQ_OP_ZONE_RESET_ALL:
>  	case REQ_OP_ZONE_OPEN:
>  	case REQ_OP_ZONE_CLOSE:
>  	case REQ_OP_ZONE_FINISH:
> -		return null_zone_mgmt(cmd, op, sector);
> +		sts = null_zone_mgmt(cmd, op, sector);
> +		break;
>  	default:
> -		return null_process_cmd(cmd, op, sector, nr_sectors);
> +		sts = null_process_cmd(cmd, op, sector, nr_sectors);
>  	}
> +	spin_unlock_irq(&dev->zone_lock);
> +
> +	return sts;
>  }
> 
> 
> 


-- 
Damien Le Moal
Western Digital Research