[dpdk-dev] [PATCH v5 11/14] eventdev: move timer adapters memory to hugepage

[Carrillo, Erik G]

Hi Pavan and Jerin,

> -----Original Message-----
> From: pbhagavatula at marvell.com <pbhagavatula at marvell.com>
> Sent: Monday, October 18, 2021 6:36 PM
> To: jerinj at marvell.com; Carrillo, Erik G <erik.g.carrillo at intel.com>
> Cc: dev at dpdk.org; Pavan Nikhilesh <pbhagavatula at marvell.com>
> Subject: [dpdk-dev] [PATCH v5 11/14] eventdev: move timer adapters
> memory to hugepage
> 
> From: Pavan Nikhilesh <pbhagavatula at marvell.com>
> 
> Move memory used by timer adapters to hugepage.
> Allocate memory on the first adapter create or lookup to address both
> primary and secondary process usecases.
> This will prevent TLB misses if any and aligns to memory structure of other
> subsystems.
> 
> Signed-off-by: Pavan Nikhilesh <pbhagavatula at marvell.com>
> ---
>  doc/guides/rel_notes/release_21_11.rst |  2 ++
> lib/eventdev/rte_event_timer_adapter.c | 36
> ++++++++++++++++++++++++--
>  2 files changed, 36 insertions(+), 2 deletions(-)
> 
> diff --git a/doc/guides/rel_notes/release_21_11.rst
> b/doc/guides/rel_notes/release_21_11.rst
> index 6442c79977..9694b32002 100644
> --- a/doc/guides/rel_notes/release_21_11.rst
> +++ b/doc/guides/rel_notes/release_21_11.rst
> @@ -226,6 +226,8 @@ API Changes
>    the crypto/security operation. This field will be used to communicate
>    events such as soft expiry with IPsec in lookaside mode.
> 
> +* eventdev: Move memory used by timer adapters to hugepage. This will
> +prevent
> +  TLB misses if any and aligns to memory structure of other subsystems.
> 
>  ABI Changes
>  -----------
> diff --git a/lib/eventdev/rte_event_timer_adapter.c
> b/lib/eventdev/rte_event_timer_adapter.c
> index ae55407042..894f532ef0 100644
> --- a/lib/eventdev/rte_event_timer_adapter.c
> +++ b/lib/eventdev/rte_event_timer_adapter.c
> @@ -33,7 +33,7 @@ RTE_LOG_REGISTER_SUFFIX(evtim_logtype,
> adapter.timer, NOTICE);
> RTE_LOG_REGISTER_SUFFIX(evtim_buffer_logtype, adapter.timer, NOTICE);
> RTE_LOG_REGISTER_SUFFIX(evtim_svc_logtype, adapter.timer.svc,
> NOTICE);
> 
> -static struct rte_event_timer_adapter
> adapters[RTE_EVENT_TIMER_ADAPTER_NUM_MAX];
> +static struct rte_event_timer_adapter *adapters;
> 
>  static const struct event_timer_adapter_ops swtim_ops;
> 
> @@ -138,6 +138,17 @@ rte_event_timer_adapter_create_ext(
>  	int n, ret;
>  	struct rte_eventdev *dev;
> 
> +	if (adapters == NULL) {
> +		adapters = rte_zmalloc("Eventdev",
> +				       sizeof(struct rte_event_timer_adapter) *
> +
> RTE_EVENT_TIMER_ADAPTER_NUM_MAX,
> +				       RTE_CACHE_LINE_SIZE);
> +		if (adapters == NULL) {
> +			rte_errno = ENOMEM;
> +			return NULL;
> +		}
> +	}
> +
>  	if (conf == NULL) {
>  		rte_errno = EINVAL;
>  		return NULL;
> @@ -312,6 +323,17 @@ rte_event_timer_adapter_lookup(uint16_t
> adapter_id)
>  	int ret;
>  	struct rte_eventdev *dev;
> 
> +	if (adapters == NULL) {
> +		adapters = rte_zmalloc("Eventdev",
> +				       sizeof(struct rte_event_timer_adapter) *
> +
> RTE_EVENT_TIMER_ADAPTER_NUM_MAX,
> +				       RTE_CACHE_LINE_SIZE);
> +		if (adapters == NULL) {
> +			rte_errno = ENOMEM;
> +			return NULL;
> +		}
> +	}
> +
>  	if (adapters[adapter_id].allocated)
>  		return &adapters[adapter_id]; /* Adapter is already loaded
> */
> 
> @@ -358,7 +380,7 @@ rte_event_timer_adapter_lookup(uint16_t
> adapter_id)  int  rte_event_timer_adapter_free(struct
> rte_event_timer_adapter *adapter)  {
> -	int ret;
> +	int i, ret;
> 
>  	ADAPTER_VALID_OR_ERR_RET(adapter, -EINVAL);
>  	FUNC_PTR_OR_ERR_RET(adapter->ops->uninit, -EINVAL); @@ -
> 382,6 +404,16 @@ rte_event_timer_adapter_free(struct
> rte_event_timer_adapter *adapter)
>  	adapter->data = NULL;
>  	adapter->allocated = 0;
> 
> +	ret = 0;
> +	for (i = 0; i < RTE_EVENT_TIMER_ADAPTER_NUM_MAX; i++)
> +		if (adapters[i].allocated)
> +			ret = adapter[i].allocated;
> +

I found a typo here, but it looks like this series has already been accepted, so I submitted the following patch for the issue:

http://patchwork.dpdk.org/project/dpdk/patch/20211020202021.1205135-1-erik.g.carrillo@intel.com/

Besides that, this patch and the others I was copied on look good to me.

Thanks,
Erik

> +	if (!ret) {
> +		rte_free(adapters);
> +		adapters = NULL;
> +	}
> +
>  	rte_eventdev_trace_timer_adapter_free(adapter);
>  	return 0;
>  }
> --
> 2.17.1