[PATCH 0/2] null_blk: zone resource management fixes

[PATCH 0/2] null_blk: zone resource management fixes

[Niklas Cassel]

Patch 1 fixes writing, opening and finishing a closed zone when both the
open and the active zone limits are reached: null_blk fails the request
instead of closing an implicitly open zone, as ZBC-3 and NVMe ZNS
require.

Patch 2 fixes zone_readonly and zone_offline, which can close an
unrelated zone, or leak the closed zone count.

Both issues were reproduced, and the fixes tested, in QEMU, with
null_blk devices created through configfs (memory_backed=0 and 1,
CONFIG_PROVE_LOCKING=y).

Developed with the help of an AI coding assistant.

Niklas Cassel (2):
  null_blk: fix opening closed zones at the active zone limit
  null_blk: don't finish zones to set them read-only or offline

 drivers/block/null_blk/zoned.c | 48 ++++++++++++++++++++++++++--------
 1 file changed, 37 insertions(+), 11 deletions(-)


base-commit: f47d223627f80cb7a1eb14c585fe6cb35de5e984
-- 
2.56.0

[PATCH 1/2] null_blk: fix opening closed zones at the active zone limit

[Niklas Cassel]

When the open zone limit is reached, null_check_open() closes an
implicitly open zone only if the active zone limit is not reached.

Closing an implicitly open zone does not change the number of active
zones, so this check is wrong. For an empty zone, the active zone limit
was already checked by null_check_zone_resources(). For a closed zone,
the check makes writing to, opening or finishing the zone fail with
BLK_STS_ZONE_OPEN_RESOURCE at the active zone limit, whereas ZBC-3
(4.5.3.2.7.2) and NVMe ZNS (2.1.1.4.1) require closing an implicitly
open zone instead.

Remove the check, and update the comment above
null_check_zone_resources() accordingly.

Fixes: dc4d137ee3b7 ("null_blk: add support for max open/active zone limit for zoned devices")
Signed-off-by: Niklas Cassel <cassel@kernel.org>
---
 drivers/block/null_blk/zoned.c | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/drivers/block/null_blk/zoned.c b/drivers/block/null_blk/zoned.c
index 7e9b5ffb9e84..ecdc8d68ca2a 100644
--- a/drivers/block/null_blk/zoned.c
+++ b/drivers/block/null_blk/zoned.c
@@ -311,10 +311,8 @@ static blk_status_t null_check_open(struct nullb_device *dev)
 		return BLK_STS_OK;
 
 	if (dev->nr_zones_imp_open) {
-		if (null_check_active(dev) == BLK_STS_OK) {
-			null_close_imp_open_zone(dev);
-			return BLK_STS_OK;
-		}
+		null_close_imp_open_zone(dev);
+		return BLK_STS_OK;
 	}
 
 	return BLK_STS_ZONE_OPEN_RESOURCE;
@@ -330,8 +328,11 @@ static blk_status_t null_check_open(struct nullb_device *dev)
  *
  * ZBC states that an implicit open zone shall be closed only if there is not
  * room within the open limit. However, with the addition of an active limit,
- * it is not certain that closing an implicit open zone will allow a new zone
- * to be opened, since we might already be at the active limit capacity.
+ * it is not certain that closing an implicit open zone will allow an empty
+ * zone to be opened, since we might already be at the active limit capacity.
+ * Therefore, the active limit is checked first for an empty zone. A closed
+ * zone already holds an active zone resource, so only the open limit has to
+ * be checked for a closed zone.
  */
 static blk_status_t null_check_zone_resources(struct nullb_device *dev,
 					      struct nullb_zone *zone)
-- 
2.56.0

Re: [PATCH 1/2] null_blk: fix opening closed zones at the active zone limit

[Damien Le Moal]

On 2026/10/21 22:21, Niklas Cassel wrote:
> When the open zone limit is reached, null_check_open() closes an
> implicitly open zone only if the active zone limit is not reached.
> 
> Closing an implicitly open zone does not change the number of active
> zones, so this check is wrong. For an empty zone, the active zone limit
> was already checked by null_check_zone_resources(). For a closed zone,
> the check makes writing to, opening or finishing the zone fail with
> BLK_STS_ZONE_OPEN_RESOURCE at the active zone limit, whereas ZBC-3
> (4.5.3.2.7.2) and NVMe ZNS (2.1.1.4.1) require closing an implicitly
> open zone instead.
> 
> Remove the check, and update the comment above
> null_check_zone_resources() accordingly.
> 
> Fixes: dc4d137ee3b7 ("null_blk: add support for max open/active zone limit for zoned devices")
> Signed-off-by: Niklas Cassel <cassel@kernel.org>

Looks good.

Reviewed-by: Damien Le Moal <dlemoal@kernel.org>

-- 
Damien Le Moal
Western Digital Research

[PATCH 2/2] null_blk: don't finish zones to set them read-only or offline

[Niklas Cassel]

null_set_zone_cond() calls null_finish_zone() to release the zone
resources held by a zone before setting it read-only or offline.

For an empty or closed zone, null_finish_zone() first runs the zone
resource checks, which can close an unrelated implicitly open zone, or
fail. As the error is ignored, a failure for a closed zone leaks
nr_zones_closed, permanently lowering the effective active zone limit.

Setting a zone read-only or offline requires no zone resources, so
release the zone resources held by the zone directly instead, using a
new helper, null_release_zone_resources().

Fixes: d3a5738849e0 ("null_blk: support read-only and offline zone conditions")
Signed-off-by: Niklas Cassel <cassel@kernel.org>
---
 drivers/block/null_blk/zoned.c | 35 +++++++++++++++++++++++++++++-----
 1 file changed, 30 insertions(+), 5 deletions(-)

diff --git a/drivers/block/null_blk/zoned.c b/drivers/block/null_blk/zoned.c
index ecdc8d68ca2a..015f5d304c7c 100644
--- a/drivers/block/null_blk/zoned.c
+++ b/drivers/block/null_blk/zoned.c
@@ -743,6 +743,33 @@ blk_status_t null_process_zoned_cmd(struct nullb_cmd *cmd, enum req_op op,
 	}
 }
 
+/*
+ * Release the zone resources held by a zone, before changing the condition of
+ * the zone to a condition that does not hold any zone resources.
+ */
+static void null_release_zone_resources(struct nullb_device *dev,
+					struct nullb_zone *zone)
+{
+	if (!dev->need_zone_res_mgmt)
+		return;
+
+	spin_lock(&dev->zone_res_lock);
+	switch (zone->cond) {
+	case BLK_ZONE_COND_IMP_OPEN:
+		dev->nr_zones_imp_open--;
+		break;
+	case BLK_ZONE_COND_EXP_OPEN:
+		dev->nr_zones_exp_open--;
+		break;
+	case BLK_ZONE_COND_CLOSED:
+		dev->nr_zones_closed--;
+		break;
+	default:
+		break;
+	}
+	spin_unlock(&dev->zone_res_lock);
+}
+
 /*
  * Set a zone in the read-only or offline condition.
  */
@@ -759,8 +786,8 @@ static void null_set_zone_cond(struct nullb_device *dev,
 	 * If the read-only condition is requested again to zones already in
 	 * read-only condition, restore back normal empty condition. Do the same
 	 * if the offline condition is requested for offline zones. Otherwise,
-	 * set the specified zone condition to the zones. Finish the zones
-	 * beforehand to free up zone resources.
+	 * set the specified zone condition to the zones. Release the zone
+	 * resources held by the zones beforehand.
 	 */
 	if (zone->cond == cond) {
 		zone->cond = BLK_ZONE_COND_EMPTY;
@@ -768,9 +795,7 @@ static void null_set_zone_cond(struct nullb_device *dev,
 		if (dev->memory_backed)
 			null_handle_discard(dev, zone->start, zone->len);
 	} else {
-		if (zone->cond != BLK_ZONE_COND_READONLY &&
-		    zone->cond != BLK_ZONE_COND_OFFLINE)
-			null_finish_zone(dev, zone);
+		null_release_zone_resources(dev, zone);
 		zone->cond = cond;
 		zone->wp = NULL_ZONE_INVALID_WP;
 	}
-- 
2.56.0

Re: [PATCH 2/2] null_blk: don't finish zones to set them read-only or offline

[Damien Le Moal]

On 2026/10/21 22:21, Niklas Cassel wrote:
> null_set_zone_cond() calls null_finish_zone() to release the zone
> resources held by a zone before setting it read-only or offline.
> 
> For an empty or closed zone, null_finish_zone() first runs the zone
> resource checks, which can close an unrelated implicitly open zone, or
> fail. As the error is ignored, a failure for a closed zone leaks
> nr_zones_closed, permanently lowering the effective active zone limit.
> 
> Setting a zone read-only or offline requires no zone resources, so
> release the zone resources held by the zone directly instead, using a
> new helper, null_release_zone_resources().
> 
> Fixes: d3a5738849e0 ("null_blk: support read-only and offline zone conditions")
> Signed-off-by: Niklas Cassel <cassel@kernel.org>

Looks good, modulo one nit below.

Reviewed-by: Damien Le Moal <dlemoal@kernel.org>

> ---
>  drivers/block/null_blk/zoned.c | 35 +++++++++++++++++++++++++++++-----
>  1 file changed, 30 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/block/null_blk/zoned.c b/drivers/block/null_blk/zoned.c
> index ecdc8d68ca2a..015f5d304c7c 100644
> --- a/drivers/block/null_blk/zoned.c
> +++ b/drivers/block/null_blk/zoned.c
> @@ -743,6 +743,33 @@ blk_status_t null_process_zoned_cmd(struct nullb_cmd *cmd, enum req_op op,
>  	}
>  }
>  
> +/*
> + * Release the zone resources held by a zone, before changing the condition of
> + * the zone to a condition that does not hold any zone resources.
> + */
> +static void null_release_zone_resources(struct nullb_device *dev,
> +					struct nullb_zone *zone)
> +{
> +	if (!dev->need_zone_res_mgmt)
> +		return;
> +
> +	spin_lock(&dev->zone_res_lock);
> +	switch (zone->cond) {
> +	case BLK_ZONE_COND_IMP_OPEN:
> +		dev->nr_zones_imp_open--;
> +		break;
> +	case BLK_ZONE_COND_EXP_OPEN:
> +		dev->nr_zones_exp_open--;
> +		break;
> +	case BLK_ZONE_COND_CLOSED:
> +		dev->nr_zones_closed--;
> +		break;
> +	default:
> +		break;
> +	}
> +	spin_unlock(&dev->zone_res_lock);
> +}
> +
>  /*
>   * Set a zone in the read-only or offline condition.
>   */
> @@ -759,8 +786,8 @@ static void null_set_zone_cond(struct nullb_device *dev,
>  	 * If the read-only condition is requested again to zones already in
>  	 * read-only condition, restore back normal empty condition. Do the same
>  	 * if the offline condition is requested for offline zones. Otherwise,
> -	 * set the specified zone condition to the zones. Finish the zones
> -	 * beforehand to free up zone resources.
> +	 * set the specified zone condition to the zones. Release the zone
> +	 * resources held by the zones beforehand.
>  	 */
>  	if (zone->cond == cond) {
>  		zone->cond = BLK_ZONE_COND_EMPTY;
> @@ -768,9 +795,7 @@ static void null_set_zone_cond(struct nullb_device *dev,

This function is called only from zone_cond_store() with cond either
BLK_ZONE_COND_READONLY or BLK_ZONE_COND_OFFLINE. Since your change is valid only
for these 2 conditions, we really should check that cond is either of the
already used values and add a comment, to avoid invalid use of
null_set_zone_cond() in the future (without tweaking it).

>  		if (dev->memory_backed)
>  			null_handle_discard(dev, zone->start, zone->len);
>  	} else {
> -		if (zone->cond != BLK_ZONE_COND_READONLY &&
> -		    zone->cond != BLK_ZONE_COND_OFFLINE)
> -			null_finish_zone(dev, zone);
> +		null_release_zone_resources(dev, zone);
>  		zone->cond = cond;
>  		zone->wp = NULL_ZONE_INVALID_WP;
>  	}


-- 
Damien Le Moal
Western Digital Research

Re: [PATCH 2/2] null_blk: don't finish zones to set them read-only or offline

[Niklas Cassel]

On Thu, Oct 22, 2026 at 09:15:32AM +0200, Damien Le Moal wrote:
> On 2026/10/21 22:21, Niklas Cassel wrote:
> > null_set_zone_cond() calls null_finish_zone() to release the zone
> > resources held by a zone before setting it read-only or offline.
> > 
> > For an empty or closed zone, null_finish_zone() first runs the zone
> > resource checks, which can close an unrelated implicitly open zone, or
> > fail. As the error is ignored, a failure for a closed zone leaks
> > nr_zones_closed, permanently lowering the effective active zone limit.
> > 
> > Setting a zone read-only or offline requires no zone resources, so
> > release the zone resources held by the zone directly instead, using a
> > new helper, null_release_zone_resources().
> > 
> > Fixes: d3a5738849e0 ("null_blk: support read-only and offline zone conditions")
> > Signed-off-by: Niklas Cassel <cassel@kernel.org>
> 
> Looks good, modulo one nit below.
> 
> Reviewed-by: Damien Le Moal <dlemoal@kernel.org>

(snip)

> > @@ -768,9 +795,7 @@ static void null_set_zone_cond(struct nullb_device *dev,
> 
> This function is called only from zone_cond_store() with cond either
> BLK_ZONE_COND_READONLY or BLK_ZONE_COND_OFFLINE. Since your change is valid only
> for these 2 conditions, we really should check that cond is either of the
> already used values and add a comment, to avoid invalid use of
> null_set_zone_cond() in the future (without tweaking it).

Could you clarify with code what you mean?


For a comment, there is one already:
https://github.com/torvalds/linux/blob/v7.3/drivers/block/null_blk/zoned.c#L746

For a check that cond is either read-only or offline, there such a check
already:
https://github.com/torvalds/linux/blob/v7.3/drivers/block/null_blk/zoned.c#L751-L752


Kind regards,
Niklas

Re: [PATCH 2/2] null_blk: don't finish zones to set them read-only or offline

[Damien Le Moal]

On 2026/10/22 9:36, Niklas Cassel wrote:
> On Thu, Oct 22, 2026 at 09:15:32AM +0200, Damien Le Moal wrote:
>> On 2026/10/21 22:21, Niklas Cassel wrote:
>>> null_set_zone_cond() calls null_finish_zone() to release the zone
>>> resources held by a zone before setting it read-only or offline.
>>>
>>> For an empty or closed zone, null_finish_zone() first runs the zone
>>> resource checks, which can close an unrelated implicitly open zone, or
>>> fail. As the error is ignored, a failure for a closed zone leaks
>>> nr_zones_closed, permanently lowering the effective active zone limit.
>>>
>>> Setting a zone read-only or offline requires no zone resources, so
>>> release the zone resources held by the zone directly instead, using a
>>> new helper, null_release_zone_resources().
>>>
>>> Fixes: d3a5738849e0 ("null_blk: support read-only and offline zone conditions")
>>> Signed-off-by: Niklas Cassel <cassel@kernel.org>
>>
>> Looks good, modulo one nit below.
>>
>> Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
> 
> (snip)
> 
>>> @@ -768,9 +795,7 @@ static void null_set_zone_cond(struct nullb_device *dev,
>>
>> This function is called only from zone_cond_store() with cond either
>> BLK_ZONE_COND_READONLY or BLK_ZONE_COND_OFFLINE. Since your change is valid only
>> for these 2 conditions, we really should check that cond is either of the
>> already used values and add a comment, to avoid invalid use of
>> null_set_zone_cond() in the future (without tweaking it).
> 
> Could you clarify with code what you mean?
> 
> 
> For a comment, there is one already:
> https://github.com/torvalds/linux/blob/v7.3/drivers/block/null_blk/zoned.c#L746
> 
> For a check that cond is either read-only or offline, there such a check
> already:
> https://github.com/torvalds/linux/blob/v7.3/drivers/block/null_blk/zoned.c#L751-L752

Ah! I missed that! So please ignore my nit comment.

That said, adding a cleanup patch that changes null_set_zone_cond() to something
like:

static void null_set_zone_readonly_or_offline(struct nullb_device *dev,
			       		      struct nullb_zone *zone,
					      bool set_offline)

would make things a lot clearer :)


-- 
Damien Le Moal
Western Digital Research