On Tue, Jan 21, 2020 at 09:13:48AM +0000, Slava Ovsiienko wrote:
> > -----Original Message-----
> > From: dev <dev-boun...@dpdk.org> On Behalf Of Slava Ovsiienko
> > Sent: Tuesday, January 21, 2020 10:24
> > To: Olivier Matz <olivier.m...@6wind.com>
> > Cc: Stephen Hemminger <step...@networkplumber.org>; dev@dpdk.org;
> > Matan Azrad <ma...@mellanox.com>; Raslan Darawsheh
> > <rasl...@mellanox.com>; Ori Kam <or...@mellanox.com>; Shahaf Shuler
> > <shah...@mellanox.com>; tho...@mellanox.net
> > Subject: Re: [dpdk-dev] [PATCH v5 1/5] mbuf: introduce routine to get
> > private
> > mbuf pool flags
> >
> > > -----Original Message-----
> > > From: Olivier Matz <olivier.m...@6wind.com>
> > > Sent: Tuesday, January 21, 2020 10:14
> > > To: Slava Ovsiienko <viachesl...@mellanox.com>
> > > Cc: Stephen Hemminger <step...@networkplumber.org>; dev@dpdk.org;
> > > Matan Azrad <ma...@mellanox.com>; Raslan Darawsheh
> > > <rasl...@mellanox.com>; Ori Kam <or...@mellanox.com>; Shahaf Shuler
> > > <shah...@mellanox.com>; tho...@mellanox.net
> > > Subject: Re: [PATCH v5 1/5] mbuf: introduce routine to get private
> > > mbuf pool flags
> > >
> > > On Tue, Jan 21, 2020 at 08:00:17AM +0000, Slava Ovsiienko wrote:
> > > > > -----Original Message-----
> > > > > From: Stephen Hemminger <step...@networkplumber.org>
> > > > > Sent: Monday, January 20, 2020 22:44
> > > > > To: Slava Ovsiienko <viachesl...@mellanox.com>
> > > > > Cc: dev@dpdk.org; Matan Azrad <ma...@mellanox.com>; Raslan
> > > Darawsheh
> > > > > <rasl...@mellanox.com>; Ori Kam <or...@mellanox.com>; Shahaf
> > > > > Shuler <shah...@mellanox.com>; olivier.m...@6wind.com;
> > > tho...@mellanox.net
> > > > > Subject: Re: [PATCH v5 1/5] mbuf: introduce routine to get private
> > > > > mbuf pool flags
> > > > >
> > > > > On Mon, 20 Jan 2020 17:23:19 +0000 Viacheslav Ovsiienko
> > > > > <viachesl...@mellanox.com> wrote:
> > > > >
> > > > > > The routine rte_pktmbuf_priv_flags is introduced to fetch the
> > > > > > flags from the mbuf memory pool private structure in unified
> > > > > > fashion.
> > > > > >
> > > > > > Signed-off-by: Viacheslav Ovsiienko <viachesl...@mellanox.com>
> > > > > > Acked-by: Olivier Matz <olivier.m...@6wind.com>
> > > > > > ---
> > > > > > lib/librte_mbuf/rte_mbuf.h | 17 +++++++++++++++++
> > > > > > 1 file changed, 17 insertions(+)
> > > > > >
> > > > > > diff --git a/lib/librte_mbuf/rte_mbuf.h
> > > > > > b/lib/librte_mbuf/rte_mbuf.h index 2d4bda2..9b0691d 100644
> > > > > > --- a/lib/librte_mbuf/rte_mbuf.h
> > > > > > +++ b/lib/librte_mbuf/rte_mbuf.h
> > > > > > @@ -306,6 +306,23 @@ struct rte_pktmbuf_pool_private {
> > > > > > uint32_t flags; /**< reserved for future use. */ };
> > > > > >
> > > > > > +/**
> > > > > > + * Return the flags from private data in an mempool structure.
> > > > > > + *
> > > > > > + * @param mp
> > > > > > + * A pointer to the mempool structure.
> > > > > > + * @return
> > > > > > + * The flags from the private data structure.
> > > > > > + */
> > > > > > +static inline uint32_t
> > > > > > +rte_pktmbuf_priv_flags(struct rte_mempool *mp) {
> > > > > > + struct rte_pktmbuf_pool_private *mbp_priv;
> > > > > > +
> > > > > > + mbp_priv = (struct rte_pktmbuf_pool_private
> > > > > *)rte_mempool_get_priv(mp);
> > > > > > + return mbp_priv->flags;
> > > > > > +}
> > > > > > +
> > > > > > #ifdef RTE_LIBRTE_MBUF_DEBUG
> > > > > >
> > > > > > /** check mbuf type in debug mode */
> > > > >
> > > > > Looks fine, but a couple of minor suggestions.
> > > > >
> > > > >
> > > > > Since this doesn't modify the mbuf, the arguments should be const.
> > > > > Since rte_mempool_get_priv returns void *, the cast is unnecessary.
> > > >
> > > > rte_mempool_get_priv() does not expect "const", so adding "const" is
> > > > a bit problematic, and we should not change the
> > > > rte_mempool_get_priv()
> > > prototype.
> > > > Do you think we should take private structure pointer directly from
> > > > the pool structure instead of calling rte_mempool_get_priv() ?
> > >
> > > I'm not sure it would work. The problem is that to get the priv, we do
> > > pool_ptr + offset. So if we want to remove the const, we'll have to do
> > > a cast to "unconst". Not sure it is worth doing it.
> > >
> > > Thanks
> > > Olivier
> > OK, I'll just remove not necessary (struct rte_pktmbuf_pool_private*) cast
> > and
> > will not introduce const qualifier for the parameter.
> >
> > With best regards, Slava
> I've checked the rte_mempool_get_priv() usage - in all header files there
> are the type casts. The "rte_mbuf.h" contains the rte_pktmbuf_priv_size() and
> rte_pktmbuf_data_room_size(), both provide the cast. What is the reason?
> C++ compatibility? Should we remove the cast in rte_pktmbuf_priv_flags()?
Removing the cast will certainly break C++ code using this header.
There is a similar case in commit a2ff2827dc84 ("mbuf: fix C++ build on
void pointer cast")
In my opinion it can stay as it is now.
