Re: [PATCH] ceph: optimizing cap reservation

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 




> On 24 Feb 2018, at 18:36, Chengguang Xu <cgxu519@xxxxxxxxxx> wrote:
> 
> When caps_avail_count is in a low level, most newly
> trimmed caps will probably go into ->caps_list and
> caps_avail_count will be increased. Hence after trimming,
> should recheck caps_avail_count to effectly reuse
> newly trimmed caps. Also, when releasing unnecessary
> caps follow the same rule of ceph_put_cap.
> 
> Signed-off-by: Chengguang Xu <cgxu519@xxxxxxxxxx>
> ---
> Only compile tested.
> 
> fs/ceph/caps.c | 41 +++++++++++++++++++++++++++++++++++++----
> 1 file changed, 37 insertions(+), 4 deletions(-)
> 
> diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
> index 6582c45..ba58422 100644
> --- a/fs/ceph/caps.c
> +++ b/fs/ceph/caps.c
> @@ -165,6 +165,7 @@ int ceph_reserve_caps(struct ceph_mds_client *mdsc,
> 	int have;
> 	int alloc = 0;
> 	int max_caps;
> +	int more_avail_caps;
> 	bool trimmed = false;
> 	struct ceph_mds_session *s;
> 	LIST_HEAD(newcaps);
> @@ -204,6 +205,24 @@ int ceph_reserve_caps(struct ceph_mds_client *mdsc,
> 					mutex_lock(&mdsc->mutex);
> 				}
> 				trimmed = true;
> +
> +				spin_lock(&mdsc->caps_list_lock);
> +				if (mdsc->caps_avail_count) {
> +					if (mdsc->caps_avail_count + have + alloc >= need)
> +						more_avail_caps = need - alloc - have;
> +					else
> +						more_avail_caps = mdsc->caps_avail_count;
> +
> +					i +=  more_avail_caps;
> +					have += more_avail_caps;
> +					mdsc->caps_avail_count -= more_avail_caps;
> +					mdsc->caps_reserve_count += more_avail_caps;
> +				}
> +				spin_unlock(&mdsc->caps_list_lock);
> +
> +				if (i >= need)
> +					break;
> +
> 				goto retry;
> 			} else {
> 				pr_warn("reserve caps ctx=%p ENOMEM "
> @@ -234,16 +253,30 @@ int ceph_reserve_caps(struct ceph_mds_client *mdsc,
> 	return 0;
> 
> out_nomem:
> +
> +	spin_lock(&mdsc->caps_list_lock);
> +	mdsc->caps_avail_count += have;
> +	mdsc->caps_reserve_count -= have;
> +
> 	while (!list_empty(&newcaps)) {
> 		cap = list_first_entry(&newcaps,
> 				struct ceph_cap, caps_item);
> 		list_del(&cap->caps_item);
> -		kmem_cache_free(ceph_cap_cachep, cap);
> +
> +		/*
> +		 * Keep some preallocated caps around (ceph_min_count), to
> +		 * avoid lots of free/alloc churn.
> +		 */
> +		if (mdsc->caps_avail_count >= mdsc->caps_reserve_count +
> +				      mdsc->caps_min_count) {
> +			kmem_cache_free(ceph_cap_cachep, cap);
> +		} else {
> +			mdsc->caps_avail_count++;
> +			mdsc->caps_total_count++;
> +			list_add(&cap->caps_item, &mdsc->caps_list);
> +		}
> 	}
> 
> -	spin_lock(&mdsc->caps_list_lock);
> -	mdsc->caps_avail_count += have;
> -	mdsc->caps_reserve_count -= have;
> 	BUG_ON(mdsc->caps_total_count != mdsc->caps_use_count +
> 					 mdsc->caps_reserve_count +
> 					 mdsc->caps_avail_count);
> -- 
> 1.8.3.1

Some lines exceed 80 characters. I make some modification to patch.

Regards
Yan, Zheng

> 

--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [CEPH Users]     [Ceph Large]     [Information on CEPH]     [Linux BTRFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]
  Powered by Linux