Re: [dpdk-dev] [PATCH] mempool: limit cache_size

[Zoltan Kiss <zoltan.kiss@linaro.org>]

On 18/05/15 13:41, Ananyev, Konstantin wrote:
>
>
>> -----Original Message-----
>> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Zoltan Kiss
>> Sent: Monday, May 18, 2015 1:28 PM
>> To: dev@dpdk.org
>> Subject: Re: [dpdk-dev] [PATCH] mempool: limit cache_size
>>
>> Hi,
>>
>> Any opinion on this patch?
>>
>> Regards,
>>
>> Zoltan
>>
>> On 13/05/15 19:59, Zoltan Kiss wrote:
>>> Otherwise cache_flushthresh can be bigger than n, and
>>> a consumer can starve others by keeping every element
>>> either in use or in the cache.
>>>
>>> Signed-off-by: Zoltan Kiss <zoltan.kiss@linaro.org>
>>> ---
>>>    lib/librte_mempool/rte_mempool.c | 3 ++-
>>>    lib/librte_mempool/rte_mempool.h | 2 +-
>>>    2 files changed, 3 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/lib/librte_mempool/rte_mempool.c b/lib/librte_mempool/rte_mempool.c
>>> index cf7ed76..ca6cd9c 100644
>>> --- a/lib/librte_mempool/rte_mempool.c
>>> +++ b/lib/librte_mempool/rte_mempool.c
>>> @@ -440,7 +440,8 @@ rte_mempool_xmem_create(const char *name, unsigned n, unsigned elt_size,
>>>    	mempool_list = RTE_TAILQ_CAST(rte_mempool_tailq.head, rte_mempool_list);
>>>
>>>    	/* asked cache too big */
>>> -	if (cache_size > RTE_MEMPOOL_CACHE_MAX_SIZE) {
>>> +	if (cache_size > RTE_MEMPOOL_CACHE_MAX_SIZE ||
>>> +	    (uint32_t) cache_size * CACHE_FLUSHTHRESH_MULTIPLIER > n) {
>>>    		rte_errno = EINVAL;
>>>    		return NULL;
>>>    	}
>
> Why just no 'cache_size > n' then?

The commit message says: "Otherwise cache_flushthresh can be bigger than 
n, and a consumer can starve others by keeping every element either in 
use or in the cache."

> Konstantin
>
>>> diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
>>> index 9001312..a4a9610 100644
>>> --- a/lib/librte_mempool/rte_mempool.h
>>> +++ b/lib/librte_mempool/rte_mempool.h
>>> @@ -468,7 +468,7 @@ typedef void (rte_mempool_ctor_t)(struct rte_mempool *, void *);
>>>     *   If cache_size is non-zero, the rte_mempool library will try to
>>>     *   limit the accesses to the common lockless pool, by maintaining a
>>>     *   per-lcore object cache. This argument must be lower or equal to
>>> - *   CONFIG_RTE_MEMPOOL_CACHE_MAX_SIZE. It is advised to choose
>>> + *   CONFIG_RTE_MEMPOOL_CACHE_MAX_SIZE and n / 1.5. It is advised to choose
>>>     *   cache_size to have "n modulo cache_size == 0": if this is
>>>     *   not the case, some elements will always stay in the pool and will
>>>     *   never be used. The access to the per-lcore table is of course
>>>