Re: [dpdk-dev] [PATCH v3 1/2] mbuf: add function returning default buffer address

[Yongseok Koh <yskoh@mellanox.com>]

> On Jan 11, 2019, at 3:17 AM, Andrew Rybchenko <arybchenko@solarflare.com> wrote:
> 
> Olivier, David,
> 
> could you take a look at naming suggested below and share your thoughts.
> My fear is that rte_mbuf_buf_addr() is too generic and true for direct mbuf
> only. That's why I'd like to highlight it in the function name.

Like the existing rte_mbuf_to_baddr(), it is to return the buf_addr of
the given mbuf. It doesn't matter whether the given mbuf is direct or not.
It should be used at user's discretion.

Yongseok

> On 1/11/19 2:03 PM, Yongseok Koh wrote:
>> On Fri, Jan 11, 2019 at 11:14:22AM +0300, Andrew Rybchenko wrote:
>> 
>>> On 1/10/19 9:35 PM, Yongseok Koh wrote:
>>> 
>>>> This patch introduces two new functions - rte_mbuf_buf_addr() and
>>>> rte_mbuf_data_addr_default().
>>>> 
>>>> rte_mbuf_buf_addr() reutrns the default buffer address of given mbuf which
>>>> comes after mbuf structure and private data.
>>>> 
>>>> rte_mbuf_data_addr_default() returns the default address of mbuf data
>>>> taking the headroom into account.
>>>> 
>>>> Signed-off-by: Yongseok Koh 
>>>> <yskoh@mellanox.com>
>>>> 
>>>> ---
>>>> 
>>>> v3:
>>>> * rename functions
>>>> 
>>>> v2:
>>>> * initial implementation
>>>> 
>>>>   lib/librte_mbuf/rte_mbuf.h | 43 ++++++++++++++++++++++++++++++++++++++++---
>>>>   1 file changed, 40 insertions(+), 3 deletions(-)
>>>> 
>>>> diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
>>>> index bc562dc8a9..486566fc28 100644
>>>> --- a/lib/librte_mbuf/rte_mbuf.h
>>>> +++ b/lib/librte_mbuf/rte_mbuf.h
>>>> @@ -788,8 +788,47 @@ rte_mbuf_from_indirect(struct rte_mbuf *mi)
>>>>   }
>>>>   /**
>>>> + * Return the default buffer address of the mbuf.
>>>> + *
>>>> + * @param mb
>>>> + *   The pointer to the mbuf.
>>>> + * @param mp
>>>> + *   The pointer to the mempool of the mbuf.
>>>> + * @return
>>>> + *   The pointer of the mbuf buffer.
>>>> + */
>>>> +static inline char * __rte_experimental
>>>> +rte_mbuf_buf_addr(struct rte_mbuf *mb, struct rte_mempool *mp)
>>>> 
>>> struct rte_mbuf has pool member. So, I don't understand why mp
>>> argument is required. I guess there is a reason, but it should be
>>> explained in comments. I see motivation in rte_mbuf_to_baddr()
>>> description, but rte_mbuf_buf_add() does not explain it.
>>> 
>> Well, I don't like to put same comment here and there but I'll add small comment
>> here.
>> 
>> 
>>> Also right now the function name looks like simple get accessor for
>>> buf_addr and I'd expect to seem one line implementation may be
>>> with extra debug checks: return mb->buf_addr.
>>> 
>> This func is suggested by David and Olivier because same code is being repeated
>> in multiple locations. This can be used to initialize a mbuf when mb->buf_addr is
>> null. And second use-case (this is my use-case) is to get the buf_addr without
>> accessing the mbuf struct when mempool of mbuf is known, e.g. Rx buffer
>> replenishment. It is definitely beneficial for performance, especially RISC
>> cores.
>> 
>> 
>>> May be rte_mbuf_direct_buf_addr() ?
>>> If so, similar below rte_mbuf_direct_data_addr_default().
>>> 
>> Regarding naming, people have different tastes. As it is acked by Olivier and
>> David, I'll keep the names.
>> 
> 
>> Thanks,
>> Yongseok
>> 
>> 
>>>> +{
>>>> +	char *buffer_addr;
>>>> +
>>>> +	buffer_addr = (char *)mb + sizeof(*mb) + rte_pktmbuf_priv_size(mp);
>>>> +	return buffer_addr;
>>>> +}
>>>> +
>>>> +
>>>> +/**
>>>> + * Return the default address of the beginning of the mbuf data.
>>>> + *
>>>> + * @param mb
>>>> + *   The pointer to the mbuf.
>>>> + * @return
>>>> + *   The pointer of the beginning of the mbuf data.
>>>> + */
>>>> +static inline char * __rte_experimental
>>>> +rte_mbuf_data_addr_default(struct rte_mbuf *mb)
>>>> +{
>>>> +	return rte_mbuf_buf_addr(mb, mb->pool) + RTE_PKTMBUF_HEADROOM;
>>>> +}
>>>> +
>>>> +/**
>>>>    * Return the buffer address embedded in the given mbuf.
>>>>    *
>>>> + * Note that accessing mempool pointer of a mbuf is expensive because the
>>>> + * pointer is stored in the 2nd cache line of mbuf. If mempool is known, it
>>>> + * is better not to reference the mempool pointer in mbuf but calling
>>>> + * rte_mbuf_buf_addr() would be more efficient.
>>>> + *
>>>>    * @param md
>>>>    *   The pointer to the mbuf.
>>>>    * @return
>>>> @@ -798,9 +837,7 @@ rte_mbuf_from_indirect(struct rte_mbuf *mi)
>>>>   static inline char *
>>>>   rte_mbuf_to_baddr(struct rte_mbuf *md)
>>>>   {
>>>> -	char *buffer_addr;
>>>> -	buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
>>>> -	return buffer_addr;
>>>> +	return rte_mbuf_buf_addr(md, md->pool);
>>>>   }
>>>>   /**
>>>> 
>