diff mbox series

[2/2] block: add max_active_zones to blk-sysfs

Message ID 20200616102546.491961-3-niklas.cassel@wdc.com (mailing list archive)
State Superseded
Headers show
Series Export max open zones and max active zones to sysfs | expand

Commit Message

Niklas Cassel June 16, 2020, 10:25 a.m. UTC
Add a new max_active zones definition in the sysfs documentation.
This definition will be common for all devices utilizing the zoned block
device support in the kernel.

Export max_active_zones according to this new definition for NVMe Zoned
Namespace devices, ZAC ATA devices (which are treated as SCSI devices by
the kernel), and ZBC SCSI devices.

Add the new max_active_zones struct member to the request_queue, rather
than as a queue limit, since this property cannot be split across stacking
drivers.

For SCSI devices, even though max active zones is not part of the ZBC/ZAC
spec, export max_active_zones as 0, signifying "no limit".

Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
---
 Documentation/block/queue-sysfs.rst |  7 +++++++
 block/blk-sysfs.c                   | 14 +++++++++++++-
 drivers/nvme/host/zns.c             |  1 +
 drivers/scsi/sd_zbc.c               |  1 +
 include/linux/blkdev.h              | 20 ++++++++++++++++++++
 5 files changed, 42 insertions(+), 1 deletion(-)

Comments

Javier González June 29, 2020, 7:42 p.m. UTC | #1
On 16.06.2020 12:25, Niklas Cassel wrote:
>Add a new max_active zones definition in the sysfs documentation.
>This definition will be common for all devices utilizing the zoned block
>device support in the kernel.
>
>Export max_active_zones according to this new definition for NVMe Zoned
>Namespace devices, ZAC ATA devices (which are treated as SCSI devices by
>the kernel), and ZBC SCSI devices.
>
>Add the new max_active_zones struct member to the request_queue, rather
>than as a queue limit, since this property cannot be split across stacking
>drivers.
>
>For SCSI devices, even though max active zones is not part of the ZBC/ZAC
>spec, export max_active_zones as 0, signifying "no limit".
>
>Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
>---
> Documentation/block/queue-sysfs.rst |  7 +++++++
> block/blk-sysfs.c                   | 14 +++++++++++++-
> drivers/nvme/host/zns.c             |  1 +
> drivers/scsi/sd_zbc.c               |  1 +
> include/linux/blkdev.h              | 20 ++++++++++++++++++++
> 5 files changed, 42 insertions(+), 1 deletion(-)
>
>diff --git a/Documentation/block/queue-sysfs.rst b/Documentation/block/queue-sysfs.rst
>index f01cf8530ae4..f261a5c84170 100644
>--- a/Documentation/block/queue-sysfs.rst
>+++ b/Documentation/block/queue-sysfs.rst
>@@ -117,6 +117,13 @@ Maximum number of elements in a DMA scatter/gather list with integrity
> data that will be submitted by the block layer core to the associated
> block driver.
>
>+max_active_zones (RO)
>+---------------------
>+For zoned block devices (zoned attribute indicating "host-managed" or
>+"host-aware"), the sum of zones belonging to any of the zone states:
>+EXPLICIT OPEN, IMPLICIT OPEN or CLOSED, is limited by this value.
>+If this value is 0, there is no limit.
>+
> max_open_zones (RO)
> -------------------
> For zoned block devices (zoned attribute indicating "host-managed" or
>diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
>index fa42961e9678..624bb4d85fc7 100644
>--- a/block/blk-sysfs.c
>+++ b/block/blk-sysfs.c
>@@ -310,6 +310,11 @@ static ssize_t queue_max_open_zones_show(struct request_queue *q, char *page)
> 	return queue_var_show(queue_max_open_zones(q), page);
> }
>
>+static ssize_t queue_max_active_zones_show(struct request_queue *q, char *page)
>+{
>+	return queue_var_show(queue_max_active_zones(q), page);
>+}
>+
> static ssize_t queue_nomerges_show(struct request_queue *q, char *page)
> {
> 	return queue_var_show((blk_queue_nomerges(q) << 1) |
>@@ -677,6 +682,11 @@ static struct queue_sysfs_entry queue_max_open_zones_entry = {
> 	.show = queue_max_open_zones_show,
> };
>
>+static struct queue_sysfs_entry queue_max_active_zones_entry = {
>+	.attr = {.name = "max_active_zones", .mode = 0444 },
>+	.show = queue_max_active_zones_show,
>+};
>+
> static struct queue_sysfs_entry queue_nomerges_entry = {
> 	.attr = {.name = "nomerges", .mode = 0644 },
> 	.show = queue_nomerges_show,
>@@ -776,6 +786,7 @@ static struct attribute *queue_attrs[] = {
> 	&queue_zoned_entry.attr,
> 	&queue_nr_zones_entry.attr,
> 	&queue_max_open_zones_entry.attr,
>+	&queue_max_active_zones_entry.attr,
> 	&queue_nomerges_entry.attr,
> 	&queue_rq_affinity_entry.attr,
> 	&queue_iostats_entry.attr,
>@@ -803,7 +814,8 @@ static umode_t queue_attr_visible(struct kobject *kobj, struct attribute *attr,
> 		(!q->mq_ops || !q->mq_ops->timeout))
> 			return 0;
>
>-	if (attr == &queue_max_open_zones_entry.attr &&
>+	if ((attr == &queue_max_open_zones_entry.attr ||
>+	     attr == &queue_max_active_zones_entry.attr) &&
> 	    !blk_queue_is_zoned(q))
> 		return 0;
>
>diff --git a/drivers/nvme/host/zns.c b/drivers/nvme/host/zns.c
>index af156529f3b6..502070763266 100644
>--- a/drivers/nvme/host/zns.c
>+++ b/drivers/nvme/host/zns.c
>@@ -83,6 +83,7 @@ int nvme_update_zone_info(struct gendisk *disk, struct nvme_ns *ns,
> 	q->limits.zoned = BLK_ZONED_HM;
> 	blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q);
> 	blk_queue_max_open_zones(q, le32_to_cpu(id->mor) + 1);
>+	blk_queue_max_active_zones(q, le32_to_cpu(id->mar) + 1);
> free_data:
> 	kfree(id);
> 	return status;
>diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c
>index aa3564139b40..d8b2c49d645b 100644
>--- a/drivers/scsi/sd_zbc.c
>+++ b/drivers/scsi/sd_zbc.c
>@@ -721,6 +721,7 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, unsigned char *buf)
> 		blk_queue_max_open_zones(q, 0);
> 	else
> 		blk_queue_max_open_zones(q, sdkp->zones_max_open);
>+	blk_queue_max_active_zones(q, 0);
> 	nr_zones = round_up(sdkp->capacity, zone_blocks) >> ilog2(zone_blocks);
>
> 	/* READ16/WRITE16 is mandatory for ZBC disks */
>diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
>index 2f332f00501d..3776140f8f20 100644
>--- a/include/linux/blkdev.h
>+++ b/include/linux/blkdev.h
>@@ -521,6 +521,7 @@ struct request_queue {
> 	unsigned long		*conv_zones_bitmap;
> 	unsigned long		*seq_zones_wlock;
> 	unsigned int		max_open_zones;
>+	unsigned int		max_active_zones;
> #endif /* CONFIG_BLK_DEV_ZONED */
>
> 	/*
>@@ -741,6 +742,17 @@ static inline unsigned int queue_max_open_zones(const struct request_queue *q)
> {
> 	return q->max_open_zones;
> }
>+
>+static inline void blk_queue_max_active_zones(struct request_queue *q,
>+		unsigned int max_active_zones)
>+{
>+	q->max_active_zones = max_active_zones;
>+}
>+
>+static inline unsigned int queue_max_active_zones(const struct request_queue *q)
>+{
>+	return q->max_active_zones;
>+}
> #else /* CONFIG_BLK_DEV_ZONED */
> static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
> {
>@@ -764,6 +776,14 @@ static inline unsigned int queue_max_open_zones(const struct request_queue *q)
> {
> 	return 0;
> }
>+static inline void blk_queue_max_active_zones(struct request_queue *q,
>+		unsigned int max_active_zones)
>+{
>+}
>+static inline unsigned int queue_max_active_zones(const struct request_queue *q)
>+{
>+	return 0;
>+}
> #endif /* CONFIG_BLK_DEV_ZONED */
>
> static inline bool rq_is_sync(struct request *rq)
>-- 
>2.26.2
>
>
>_______________________________________________
>linux-nvme mailing list
>linux-nvme@lists.infradead.org
>http://lists.infradead.org/mailman/listinfo/linux-nvme

Looks good to me

Reviewed-by: Javier González <javier@javigon.com>
Damien Le Moal June 30, 2020, 1:51 a.m. UTC | #2
On 2020/06/16 19:28, Niklas Cassel wrote:
> Add a new max_active zones definition in the sysfs documentation.
> This definition will be common for all devices utilizing the zoned block
> device support in the kernel.
> 
> Export max_active_zones according to this new definition for NVMe Zoned
> Namespace devices, ZAC ATA devices (which are treated as SCSI devices by
> the kernel), and ZBC SCSI devices.
> 
> Add the new max_active_zones struct member to the request_queue, rather
> than as a queue limit, since this property cannot be split across stacking
> drivers.

Same comment as for max_open_zones.

> 
> For SCSI devices, even though max active zones is not part of the ZBC/ZAC
> spec, export max_active_zones as 0, signifying "no limit".
> 
> Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
> ---
>  Documentation/block/queue-sysfs.rst |  7 +++++++
>  block/blk-sysfs.c                   | 14 +++++++++++++-
>  drivers/nvme/host/zns.c             |  1 +
>  drivers/scsi/sd_zbc.c               |  1 +
>  include/linux/blkdev.h              | 20 ++++++++++++++++++++
>  5 files changed, 42 insertions(+), 1 deletion(-)
> 
> diff --git a/Documentation/block/queue-sysfs.rst b/Documentation/block/queue-sysfs.rst
> index f01cf8530ae4..f261a5c84170 100644
> --- a/Documentation/block/queue-sysfs.rst
> +++ b/Documentation/block/queue-sysfs.rst
> @@ -117,6 +117,13 @@ Maximum number of elements in a DMA scatter/gather list with integrity
>  data that will be submitted by the block layer core to the associated
>  block driver.
>  
> +max_active_zones (RO)
> +---------------------
> +For zoned block devices (zoned attribute indicating "host-managed" or
> +"host-aware"), the sum of zones belonging to any of the zone states:
> +EXPLICIT OPEN, IMPLICIT OPEN or CLOSED, is limited by this value.
> +If this value is 0, there is no limit.
> +
>  max_open_zones (RO)
>  -------------------
>  For zoned block devices (zoned attribute indicating "host-managed" or
> diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
> index fa42961e9678..624bb4d85fc7 100644
> --- a/block/blk-sysfs.c
> +++ b/block/blk-sysfs.c
> @@ -310,6 +310,11 @@ static ssize_t queue_max_open_zones_show(struct request_queue *q, char *page)
>  	return queue_var_show(queue_max_open_zones(q), page);
>  }
>  
> +static ssize_t queue_max_active_zones_show(struct request_queue *q, char *page)
> +{
> +	return queue_var_show(queue_max_active_zones(q), page);
> +}
> +
>  static ssize_t queue_nomerges_show(struct request_queue *q, char *page)
>  {
>  	return queue_var_show((blk_queue_nomerges(q) << 1) |
> @@ -677,6 +682,11 @@ static struct queue_sysfs_entry queue_max_open_zones_entry = {
>  	.show = queue_max_open_zones_show,
>  };
>  
> +static struct queue_sysfs_entry queue_max_active_zones_entry = {
> +	.attr = {.name = "max_active_zones", .mode = 0444 },
> +	.show = queue_max_active_zones_show,
> +};
> +
>  static struct queue_sysfs_entry queue_nomerges_entry = {
>  	.attr = {.name = "nomerges", .mode = 0644 },
>  	.show = queue_nomerges_show,
> @@ -776,6 +786,7 @@ static struct attribute *queue_attrs[] = {
>  	&queue_zoned_entry.attr,
>  	&queue_nr_zones_entry.attr,
>  	&queue_max_open_zones_entry.attr,
> +	&queue_max_active_zones_entry.attr,
>  	&queue_nomerges_entry.attr,
>  	&queue_rq_affinity_entry.attr,
>  	&queue_iostats_entry.attr,
> @@ -803,7 +814,8 @@ static umode_t queue_attr_visible(struct kobject *kobj, struct attribute *attr,
>  		(!q->mq_ops || !q->mq_ops->timeout))
>  			return 0;
>  
> -	if (attr == &queue_max_open_zones_entry.attr &&
> +	if ((attr == &queue_max_open_zones_entry.attr ||
> +	     attr == &queue_max_active_zones_entry.attr) &&
>  	    !blk_queue_is_zoned(q))
>  		return 0;
>  
> diff --git a/drivers/nvme/host/zns.c b/drivers/nvme/host/zns.c
> index af156529f3b6..502070763266 100644
> --- a/drivers/nvme/host/zns.c
> +++ b/drivers/nvme/host/zns.c
> @@ -83,6 +83,7 @@ int nvme_update_zone_info(struct gendisk *disk, struct nvme_ns *ns,
>  	q->limits.zoned = BLK_ZONED_HM;
>  	blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q);
>  	blk_queue_max_open_zones(q, le32_to_cpu(id->mor) + 1);
> +	blk_queue_max_active_zones(q, le32_to_cpu(id->mar) + 1);
>  free_data:
>  	kfree(id);
>  	return status;
> diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c
> index aa3564139b40..d8b2c49d645b 100644
> --- a/drivers/scsi/sd_zbc.c
> +++ b/drivers/scsi/sd_zbc.c
> @@ -721,6 +721,7 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, unsigned char *buf)
>  		blk_queue_max_open_zones(q, 0);
>  	else
>  		blk_queue_max_open_zones(q, sdkp->zones_max_open);
> +	blk_queue_max_active_zones(q, 0);
>  	nr_zones = round_up(sdkp->capacity, zone_blocks) >> ilog2(zone_blocks);
>  
>  	/* READ16/WRITE16 is mandatory for ZBC disks */
> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> index 2f332f00501d..3776140f8f20 100644
> --- a/include/linux/blkdev.h
> +++ b/include/linux/blkdev.h
> @@ -521,6 +521,7 @@ struct request_queue {
>  	unsigned long		*conv_zones_bitmap;
>  	unsigned long		*seq_zones_wlock;
>  	unsigned int		max_open_zones;
> +	unsigned int		max_active_zones;
>  #endif /* CONFIG_BLK_DEV_ZONED */
>  
>  	/*
> @@ -741,6 +742,17 @@ static inline unsigned int queue_max_open_zones(const struct request_queue *q)
>  {
>  	return q->max_open_zones;
>  }
> +
> +static inline void blk_queue_max_active_zones(struct request_queue *q,
> +		unsigned int max_active_zones)
> +{
> +	q->max_active_zones = max_active_zones;
> +}
> +
> +static inline unsigned int queue_max_active_zones(const struct request_queue *q)
> +{
> +	return q->max_active_zones;
> +}
>  #else /* CONFIG_BLK_DEV_ZONED */
>  static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
>  {
> @@ -764,6 +776,14 @@ static inline unsigned int queue_max_open_zones(const struct request_queue *q)
>  {
>  	return 0;
>  }
> +static inline void blk_queue_max_active_zones(struct request_queue *q,
> +		unsigned int max_active_zones)
> +{
> +}

Same comment as for max_open_zones here.

> +static inline unsigned int queue_max_active_zones(const struct request_queue *q)
> +{
> +	return 0;
> +}
>  #endif /* CONFIG_BLK_DEV_ZONED */
>  
>  static inline bool rq_is_sync(struct request *rq)
>
Javier González July 1, 2020, 11:16 a.m. UTC | #3
On 16.06.2020 12:25, Niklas Cassel wrote:
>Add a new max_active zones definition in the sysfs documentation.
>This definition will be common for all devices utilizing the zoned block
>device support in the kernel.
>
>Export max_active_zones according to this new definition for NVMe Zoned
>Namespace devices, ZAC ATA devices (which are treated as SCSI devices by
>the kernel), and ZBC SCSI devices.
>
>Add the new max_active_zones struct member to the request_queue, rather
>than as a queue limit, since this property cannot be split across stacking
>drivers.
>
>For SCSI devices, even though max active zones is not part of the ZBC/ZAC
>spec, export max_active_zones as 0, signifying "no limit".
>
>Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
>---
> Documentation/block/queue-sysfs.rst |  7 +++++++
> block/blk-sysfs.c                   | 14 +++++++++++++-
> drivers/nvme/host/zns.c             |  1 +
> drivers/scsi/sd_zbc.c               |  1 +
> include/linux/blkdev.h              | 20 ++++++++++++++++++++
> 5 files changed, 42 insertions(+), 1 deletion(-)
>
>diff --git a/Documentation/block/queue-sysfs.rst b/Documentation/block/queue-sysfs.rst
>index f01cf8530ae4..f261a5c84170 100644
>--- a/Documentation/block/queue-sysfs.rst
>+++ b/Documentation/block/queue-sysfs.rst
>@@ -117,6 +117,13 @@ Maximum number of elements in a DMA scatter/gather list with integrity
> data that will be submitted by the block layer core to the associated
> block driver.
>
>+max_active_zones (RO)
>+---------------------
>+For zoned block devices (zoned attribute indicating "host-managed" or
>+"host-aware"), the sum of zones belonging to any of the zone states:
>+EXPLICIT OPEN, IMPLICIT OPEN or CLOSED, is limited by this value.
>+If this value is 0, there is no limit.
>+
> max_open_zones (RO)
> -------------------
> For zoned block devices (zoned attribute indicating "host-managed" or
>diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
>index fa42961e9678..624bb4d85fc7 100644
>--- a/block/blk-sysfs.c
>+++ b/block/blk-sysfs.c
>@@ -310,6 +310,11 @@ static ssize_t queue_max_open_zones_show(struct request_queue *q, char *page)
> 	return queue_var_show(queue_max_open_zones(q), page);
> }
>
>+static ssize_t queue_max_active_zones_show(struct request_queue *q, char *page)
>+{
>+	return queue_var_show(queue_max_active_zones(q), page);
>+}
>+
> static ssize_t queue_nomerges_show(struct request_queue *q, char *page)
> {
> 	return queue_var_show((blk_queue_nomerges(q) << 1) |
>@@ -677,6 +682,11 @@ static struct queue_sysfs_entry queue_max_open_zones_entry = {
> 	.show = queue_max_open_zones_show,
> };
>
>+static struct queue_sysfs_entry queue_max_active_zones_entry = {
>+	.attr = {.name = "max_active_zones", .mode = 0444 },
>+	.show = queue_max_active_zones_show,
>+};
>+
> static struct queue_sysfs_entry queue_nomerges_entry = {
> 	.attr = {.name = "nomerges", .mode = 0644 },
> 	.show = queue_nomerges_show,
>@@ -776,6 +786,7 @@ static struct attribute *queue_attrs[] = {
> 	&queue_zoned_entry.attr,
> 	&queue_nr_zones_entry.attr,
> 	&queue_max_open_zones_entry.attr,
>+	&queue_max_active_zones_entry.attr,
> 	&queue_nomerges_entry.attr,
> 	&queue_rq_affinity_entry.attr,
> 	&queue_iostats_entry.attr,
>@@ -803,7 +814,8 @@ static umode_t queue_attr_visible(struct kobject *kobj, struct attribute *attr,
> 		(!q->mq_ops || !q->mq_ops->timeout))
> 			return 0;
>
>-	if (attr == &queue_max_open_zones_entry.attr &&
>+	if ((attr == &queue_max_open_zones_entry.attr ||
>+	     attr == &queue_max_active_zones_entry.attr) &&
> 	    !blk_queue_is_zoned(q))
> 		return 0;
>
>diff --git a/drivers/nvme/host/zns.c b/drivers/nvme/host/zns.c
>index af156529f3b6..502070763266 100644
>--- a/drivers/nvme/host/zns.c
>+++ b/drivers/nvme/host/zns.c
>@@ -83,6 +83,7 @@ int nvme_update_zone_info(struct gendisk *disk, struct nvme_ns *ns,
> 	q->limits.zoned = BLK_ZONED_HM;
> 	blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q);
> 	blk_queue_max_open_zones(q, le32_to_cpu(id->mor) + 1);
>+	blk_queue_max_active_zones(q, le32_to_cpu(id->mar) + 1);
> free_data:
> 	kfree(id);
> 	return status;
>diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c
>index aa3564139b40..d8b2c49d645b 100644
>--- a/drivers/scsi/sd_zbc.c
>+++ b/drivers/scsi/sd_zbc.c
>@@ -721,6 +721,7 @@ int sd_zbc_read_zones(struct scsi_disk *sdkp, unsigned char *buf)
> 		blk_queue_max_open_zones(q, 0);
> 	else
> 		blk_queue_max_open_zones(q, sdkp->zones_max_open);
>+	blk_queue_max_active_zones(q, 0);
> 	nr_zones = round_up(sdkp->capacity, zone_blocks) >> ilog2(zone_blocks);
>
> 	/* READ16/WRITE16 is mandatory for ZBC disks */
>diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
>index 2f332f00501d..3776140f8f20 100644
>--- a/include/linux/blkdev.h
>+++ b/include/linux/blkdev.h
>@@ -521,6 +521,7 @@ struct request_queue {
> 	unsigned long		*conv_zones_bitmap;
> 	unsigned long		*seq_zones_wlock;
> 	unsigned int		max_open_zones;
>+	unsigned int		max_active_zones;
> #endif /* CONFIG_BLK_DEV_ZONED */

Looking a second time at these patches, wouldn't it make sense to move
this to queue_limits?

Javier
Niklas Cassel July 2, 2020, 8:41 a.m. UTC | #4
On Wed, Jul 01, 2020 at 01:16:52PM +0200, Javier González wrote:
> On 16.06.2020 12:25, Niklas Cassel wrote:
> > Add a new max_active zones definition in the sysfs documentation.
> > This definition will be common for all devices utilizing the zoned block
> > device support in the kernel.
> > 
> > Export max_active_zones according to this new definition for NVMe Zoned
> > Namespace devices, ZAC ATA devices (which are treated as SCSI devices by
> > the kernel), and ZBC SCSI devices.
> > 
> > Add the new max_active_zones struct member to the request_queue, rather
> > than as a queue limit, since this property cannot be split across stacking
> > drivers.
> > 
> > For SCSI devices, even though max active zones is not part of the ZBC/ZAC
> > spec, export max_active_zones as 0, signifying "no limit".
> > 
> > Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
> > ---

(snip)
 
> Looking a second time at these patches, wouldn't it make sense to move
> this to queue_limits?

Hello Javier,

The problem with having MAR/MOR as queue_limits, is that they
then would be split across stacking drivers/device-mapper targets.
However, MAR/MOR are not splittable, at least not the way the
block layer works today.

If the block layer and drivers ever change so that they do
accounting of zone conditions, then we could divide the MAR/MOR to
be split over stacking drivers, but because of performance reasons,
this will probably never happen.
In the unlikely event that it did happen, we would still use the
same sysfs-path for these properties, the only thing that would
change would be that these would be moved into queue_limits.


So the way the code looks right now, these properties cannot
be split, therefore I chose to put them inside request_queue
(just like nr_zones), rather than request_queue->limits
(which is of type struct queue_limits).

nr_zones is also exposed as a sysfs property, even though it
is part of request_queue, so I don't see why MAR/MOR can't do
the same. Also see Damien's replies to PATCH 1/2 of this series,
which reaches the same conclusion.


Kind regards,
Niklas
Javier González July 2, 2020, 10:20 a.m. UTC | #5
On 02.07.2020 08:41, Niklas Cassel wrote:
>On Wed, Jul 01, 2020 at 01:16:52PM +0200, Javier González wrote:
>> On 16.06.2020 12:25, Niklas Cassel wrote:
>> > Add a new max_active zones definition in the sysfs documentation.
>> > This definition will be common for all devices utilizing the zoned block
>> > device support in the kernel.
>> >
>> > Export max_active_zones according to this new definition for NVMe Zoned
>> > Namespace devices, ZAC ATA devices (which are treated as SCSI devices by
>> > the kernel), and ZBC SCSI devices.
>> >
>> > Add the new max_active_zones struct member to the request_queue, rather
>> > than as a queue limit, since this property cannot be split across stacking
>> > drivers.
>> >
>> > For SCSI devices, even though max active zones is not part of the ZBC/ZAC
>> > spec, export max_active_zones as 0, signifying "no limit".
>> >
>> > Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
>> > ---
>
>(snip)
>
>> Looking a second time at these patches, wouldn't it make sense to move
>> this to queue_limits?
>
>Hello Javier,
>
>The problem with having MAR/MOR as queue_limits, is that they
>then would be split across stacking drivers/device-mapper targets.
>However, MAR/MOR are not splittable, at least not the way the
>block layer works today.
>
>If the block layer and drivers ever change so that they do
>accounting of zone conditions, then we could divide the MAR/MOR to
>be split over stacking drivers, but because of performance reasons,
>this will probably never happen.
>In the unlikely event that it did happen, we would still use the
>same sysfs-path for these properties, the only thing that would
>change would be that these would be moved into queue_limits.
>
>
>So the way the code looks right now, these properties cannot
>be split, therefore I chose to put them inside request_queue
>(just like nr_zones), rather than request_queue->limits
>(which is of type struct queue_limits).
>
>nr_zones is also exposed as a sysfs property, even though it
>is part of request_queue, so I don't see why MAR/MOR can't do
>the same. Also see Damien's replies to PATCH 1/2 of this series,
>which reaches the same conclusion.
>

Thanks for explaining Niklas - makes sense. I just looked at your patch
again while adding other attributes and thought it would be worth asking
the reason behind it.

You can keep the reviewed-by on the 2 patches.

Javier
diff mbox series

Patch

diff --git a/Documentation/block/queue-sysfs.rst b/Documentation/block/queue-sysfs.rst
index f01cf8530ae4..f261a5c84170 100644
--- a/Documentation/block/queue-sysfs.rst
+++ b/Documentation/block/queue-sysfs.rst
@@ -117,6 +117,13 @@  Maximum number of elements in a DMA scatter/gather list with integrity
 data that will be submitted by the block layer core to the associated
 block driver.
 
+max_active_zones (RO)
+---------------------
+For zoned block devices (zoned attribute indicating "host-managed" or
+"host-aware"), the sum of zones belonging to any of the zone states:
+EXPLICIT OPEN, IMPLICIT OPEN or CLOSED, is limited by this value.
+If this value is 0, there is no limit.
+
 max_open_zones (RO)
 -------------------
 For zoned block devices (zoned attribute indicating "host-managed" or
diff --git a/block/blk-sysfs.c b/block/blk-sysfs.c
index fa42961e9678..624bb4d85fc7 100644
--- a/block/blk-sysfs.c
+++ b/block/blk-sysfs.c
@@ -310,6 +310,11 @@  static ssize_t queue_max_open_zones_show(struct request_queue *q, char *page)
 	return queue_var_show(queue_max_open_zones(q), page);
 }
 
+static ssize_t queue_max_active_zones_show(struct request_queue *q, char *page)
+{
+	return queue_var_show(queue_max_active_zones(q), page);
+}
+
 static ssize_t queue_nomerges_show(struct request_queue *q, char *page)
 {
 	return queue_var_show((blk_queue_nomerges(q) << 1) |
@@ -677,6 +682,11 @@  static struct queue_sysfs_entry queue_max_open_zones_entry = {
 	.show = queue_max_open_zones_show,
 };
 
+static struct queue_sysfs_entry queue_max_active_zones_entry = {
+	.attr = {.name = "max_active_zones", .mode = 0444 },
+	.show = queue_max_active_zones_show,
+};
+
 static struct queue_sysfs_entry queue_nomerges_entry = {
 	.attr = {.name = "nomerges", .mode = 0644 },
 	.show = queue_nomerges_show,
@@ -776,6 +786,7 @@  static struct attribute *queue_attrs[] = {
 	&queue_zoned_entry.attr,
 	&queue_nr_zones_entry.attr,
 	&queue_max_open_zones_entry.attr,
+	&queue_max_active_zones_entry.attr,
 	&queue_nomerges_entry.attr,
 	&queue_rq_affinity_entry.attr,
 	&queue_iostats_entry.attr,
@@ -803,7 +814,8 @@  static umode_t queue_attr_visible(struct kobject *kobj, struct attribute *attr,
 		(!q->mq_ops || !q->mq_ops->timeout))
 			return 0;
 
-	if (attr == &queue_max_open_zones_entry.attr &&
+	if ((attr == &queue_max_open_zones_entry.attr ||
+	     attr == &queue_max_active_zones_entry.attr) &&
 	    !blk_queue_is_zoned(q))
 		return 0;
 
diff --git a/drivers/nvme/host/zns.c b/drivers/nvme/host/zns.c
index af156529f3b6..502070763266 100644
--- a/drivers/nvme/host/zns.c
+++ b/drivers/nvme/host/zns.c
@@ -83,6 +83,7 @@  int nvme_update_zone_info(struct gendisk *disk, struct nvme_ns *ns,
 	q->limits.zoned = BLK_ZONED_HM;
 	blk_queue_flag_set(QUEUE_FLAG_ZONE_RESETALL, q);
 	blk_queue_max_open_zones(q, le32_to_cpu(id->mor) + 1);
+	blk_queue_max_active_zones(q, le32_to_cpu(id->mar) + 1);
 free_data:
 	kfree(id);
 	return status;
diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c
index aa3564139b40..d8b2c49d645b 100644
--- a/drivers/scsi/sd_zbc.c
+++ b/drivers/scsi/sd_zbc.c
@@ -721,6 +721,7 @@  int sd_zbc_read_zones(struct scsi_disk *sdkp, unsigned char *buf)
 		blk_queue_max_open_zones(q, 0);
 	else
 		blk_queue_max_open_zones(q, sdkp->zones_max_open);
+	blk_queue_max_active_zones(q, 0);
 	nr_zones = round_up(sdkp->capacity, zone_blocks) >> ilog2(zone_blocks);
 
 	/* READ16/WRITE16 is mandatory for ZBC disks */
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index 2f332f00501d..3776140f8f20 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -521,6 +521,7 @@  struct request_queue {
 	unsigned long		*conv_zones_bitmap;
 	unsigned long		*seq_zones_wlock;
 	unsigned int		max_open_zones;
+	unsigned int		max_active_zones;
 #endif /* CONFIG_BLK_DEV_ZONED */
 
 	/*
@@ -741,6 +742,17 @@  static inline unsigned int queue_max_open_zones(const struct request_queue *q)
 {
 	return q->max_open_zones;
 }
+
+static inline void blk_queue_max_active_zones(struct request_queue *q,
+		unsigned int max_active_zones)
+{
+	q->max_active_zones = max_active_zones;
+}
+
+static inline unsigned int queue_max_active_zones(const struct request_queue *q)
+{
+	return q->max_active_zones;
+}
 #else /* CONFIG_BLK_DEV_ZONED */
 static inline unsigned int blk_queue_nr_zones(struct request_queue *q)
 {
@@ -764,6 +776,14 @@  static inline unsigned int queue_max_open_zones(const struct request_queue *q)
 {
 	return 0;
 }
+static inline void blk_queue_max_active_zones(struct request_queue *q,
+		unsigned int max_active_zones)
+{
+}
+static inline unsigned int queue_max_active_zones(const struct request_queue *q)
+{
+	return 0;
+}
 #endif /* CONFIG_BLK_DEV_ZONED */
 
 static inline bool rq_is_sync(struct request *rq)