Re: [dm-devel] [PATCH 11/12] dm-zoned: round-robin load balancer for reclaiming zones
On 5/25/20 4:42 AM, Damien Le Moal wrote:
On 2020/05/23 0:39, Hannes Reinecke wrote:
When reclaiming zones we should arbitrate between the zoned
devices to get a better throughput. So implement a simple
round-robin load balancer between the zoned devices.
Signed-off-by: Hannes Reinecke
---
drivers/md/dm-zoned-metadata.c | 8 +++-
1 file changed, 7 insertions(+), 1 deletion(-)
diff --git a/drivers/md/dm-zoned-metadata.c b/drivers/md/dm-zoned-metadata.c
index 87784e7785bc..25dcad2a565f 100644
--- a/drivers/md/dm-zoned-metadata.c
+++ b/drivers/md/dm-zoned-metadata.c
@@ -171,6 +171,8 @@ struct dmz_metadata {
unsigned intnr_reserved_seq;
unsigned intnr_chunks;
+ unsigned int last_alloc_idx;
+
/* Zone information array */
struct xarray zones;
@@ -2178,7 +2180,7 @@ struct dm_zone *dmz_alloc_zone(struct dmz_metadata *zmd, unsigned long flags)
{
struct list_head *list;
struct dm_zone *zone;
- unsigned int dev_idx = 0;
+ unsigned int dev_idx = zmd->last_alloc_idx;
again:
if (flags & DMZ_ALLOC_CACHE)
@@ -2214,6 +2216,9 @@ struct dm_zone *dmz_alloc_zone(struct dmz_metadata *zmd,
unsigned long flags)
zone = list_first_entry(list, struct dm_zone, link);
list_del_init(>link);
+ if (!(flags & DMZ_ALLOC_CACHE))
+ zmd->last_alloc_idx = (dev_idx + 1) % zmd->nr_devs;
+
if (dmz_is_cache(zone))
atomic_dec(>unmap_nr_cache);
else if (dmz_is_rnd(zone))
@@ -2839,6 +2844,7 @@ int dmz_ctr_metadata(struct dmz_dev *dev, int num_dev,
zmd->dev = dev;
zmd->nr_devs = num_dev;
zmd->mblk_rbtree = RB_ROOT;
+ zmd->last_alloc_idx = 0;
init_rwsem(>mblk_sem);
mutex_init(>mblk_flush_lock);
spin_lock_init(>mblk_lock);
OK. So my comment on patch 8 is already addressed. Or at least partly... Where
is last_alloc_idx actually used ? It looks like this only sets last_alloc_idx
but do not use that value on entry to dmz_alloc_zone() to allocate the zone.
Aw, fsck. Something went astray when generating the patches.
Will be fixing it up for the next round.
Cheers,
Hannes
--
Dr. Hannes ReineckeTeamlead Storage & Networking
h...@suse.de +49 911 74053 688
SUSE Software Solutions GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), Geschäftsführer: Felix Imendörffer
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
Re: [dm-devel] [PATCH 11/12] dm-zoned: round-robin load balancer for reclaiming zones
On 2020/05/23 0:39, Hannes Reinecke wrote:
> When reclaiming zones we should arbitrate between the zoned
> devices to get a better throughput. So implement a simple
> round-robin load balancer between the zoned devices.
>
> Signed-off-by: Hannes Reinecke
> ---
> drivers/md/dm-zoned-metadata.c | 8 +++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/md/dm-zoned-metadata.c b/drivers/md/dm-zoned-metadata.c
> index 87784e7785bc..25dcad2a565f 100644
> --- a/drivers/md/dm-zoned-metadata.c
> +++ b/drivers/md/dm-zoned-metadata.c
> @@ -171,6 +171,8 @@ struct dmz_metadata {
> unsigned intnr_reserved_seq;
> unsigned intnr_chunks;
>
> + unsigned intlast_alloc_idx;
> +
> /* Zone information array */
> struct xarray zones;
>
> @@ -2178,7 +2180,7 @@ struct dm_zone *dmz_alloc_zone(struct dmz_metadata
> *zmd, unsigned long flags)
> {
> struct list_head *list;
> struct dm_zone *zone;
> - unsigned int dev_idx = 0;
> + unsigned int dev_idx = zmd->last_alloc_idx;
>
> again:
> if (flags & DMZ_ALLOC_CACHE)
> @@ -2214,6 +2216,9 @@ struct dm_zone *dmz_alloc_zone(struct dmz_metadata
> *zmd, unsigned long flags)
> zone = list_first_entry(list, struct dm_zone, link);
> list_del_init(>link);
>
> + if (!(flags & DMZ_ALLOC_CACHE))
> + zmd->last_alloc_idx = (dev_idx + 1) % zmd->nr_devs;
> +
> if (dmz_is_cache(zone))
> atomic_dec(>unmap_nr_cache);
> else if (dmz_is_rnd(zone))
> @@ -2839,6 +2844,7 @@ int dmz_ctr_metadata(struct dmz_dev *dev, int num_dev,
> zmd->dev = dev;
> zmd->nr_devs = num_dev;
> zmd->mblk_rbtree = RB_ROOT;
> + zmd->last_alloc_idx = 0;
> init_rwsem(>mblk_sem);
> mutex_init(>mblk_flush_lock);
> spin_lock_init(>mblk_lock);
>
OK. So my comment on patch 8 is already addressed. Or at least partly... Where
is last_alloc_idx actually used ? It looks like this only sets last_alloc_idx
but do not use that value on entry to dmz_alloc_zone() to allocate the zone.
--
Damien Le Moal
Western Digital Research
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
[dm-devel] [PATCH 11/12] dm-zoned: round-robin load balancer for reclaiming zones
When reclaiming zones we should arbitrate between the zoned
devices to get a better throughput. So implement a simple
round-robin load balancer between the zoned devices.
Signed-off-by: Hannes Reinecke
---
drivers/md/dm-zoned-metadata.c | 8 +++-
1 file changed, 7 insertions(+), 1 deletion(-)
diff --git a/drivers/md/dm-zoned-metadata.c b/drivers/md/dm-zoned-metadata.c
index 87784e7785bc..25dcad2a565f 100644
--- a/drivers/md/dm-zoned-metadata.c
+++ b/drivers/md/dm-zoned-metadata.c
@@ -171,6 +171,8 @@ struct dmz_metadata {
unsigned intnr_reserved_seq;
unsigned intnr_chunks;
+ unsigned intlast_alloc_idx;
+
/* Zone information array */
struct xarray zones;
@@ -2178,7 +2180,7 @@ struct dm_zone *dmz_alloc_zone(struct dmz_metadata *zmd,
unsigned long flags)
{
struct list_head *list;
struct dm_zone *zone;
- unsigned int dev_idx = 0;
+ unsigned int dev_idx = zmd->last_alloc_idx;
again:
if (flags & DMZ_ALLOC_CACHE)
@@ -2214,6 +2216,9 @@ struct dm_zone *dmz_alloc_zone(struct dmz_metadata *zmd,
unsigned long flags)
zone = list_first_entry(list, struct dm_zone, link);
list_del_init(>link);
+ if (!(flags & DMZ_ALLOC_CACHE))
+ zmd->last_alloc_idx = (dev_idx + 1) % zmd->nr_devs;
+
if (dmz_is_cache(zone))
atomic_dec(>unmap_nr_cache);
else if (dmz_is_rnd(zone))
@@ -2839,6 +2844,7 @@ int dmz_ctr_metadata(struct dmz_dev *dev, int num_dev,
zmd->dev = dev;
zmd->nr_devs = num_dev;
zmd->mblk_rbtree = RB_ROOT;
+ zmd->last_alloc_idx = 0;
init_rwsem(>mblk_sem);
mutex_init(>mblk_flush_lock);
spin_lock_init(>mblk_lock);
--
2.16.4
--
dm-devel mailing list
dm-devel@redhat.com
https://www.redhat.com/mailman/listinfo/dm-devel
