On Mon, Mar 30, 2015 at 10:04:20PM +0300, Vadim Suraev wrote:
> Hi, Neil
>
> >I think what you need to do here is enhance the underlying pktmbuf
> interface
> >such that an rte_mbuf structure has a destructor method association with it
> >which is called when its refcnt reaches zero. That way the
> >rte_pktmbuf_bulk_free function can just decrement the refcnt on each
> >mbuf_structure, and the pool as a whole can be returned when the destructor
> >function discovers that all mbufs in that bulk pool are freed.
>
> I thought again and it looks to me that if mempool_cache is enabled,
> rte_pktmbuf_bulk_free and are redundant because the logic would be very
> similar to already implemented in rte_mempool. Probably the only
> rte_pktmbuf_alloc_bulk makes sense in this patch?
>
> Regards,
> Vadim.
>
Looking at it, yes, I agree, using an externally allocated large contiguous
block of memory, mapped with rte_mempool_xmem_create, then allocating with
rte_pktmbuf_alloc would likely work in exactly the same way. I'd argue that
even the bulk alloc function isn't really needed, as its implementation seems
like it would just be a for loop with 2-3 lines in it.
Neil
> On Wed, Mar 18, 2015 at 10:58 PM, Neil Horman <nhorman at tuxdriver.com>
> wrote:
>
> > On Wed, Mar 18, 2015 at 10:21:18PM +0200, vadim.suraev at gmail.com wrote:
> > > From: "vadim.suraev at gmail.com" <vadim.suraev at gmail.com>
> > >
> > > This patch adds mbuf bulk allocation/freeing functions and unittest
> > >
> > > Signed-off-by: Vadim Suraev
> > > <vadim.suraev at gmail.com>
> > > ---
> > > New in v2:
> > > - function rte_pktmbuf_alloc_bulk added
> > > - function rte_pktmbuf_bulk_free added
> > > - function rte_pktmbuf_free_chain added
> > > - applied reviewers' comments
> > >
> > > app/test/test_mbuf.c | 94
> > +++++++++++++++++++++++++++++++++++++++++++-
> > > lib/librte_mbuf/rte_mbuf.h | 91
> > ++++++++++++++++++++++++++++++++++++++++++
> > > 2 files changed, 184 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/app/test/test_mbuf.c b/app/test/test_mbuf.c
> > > index 1ff66cb..b20c6a4 100644
> > > --- a/app/test/test_mbuf.c
> > > +++ b/app/test/test_mbuf.c
> > > @@ -77,6 +77,7 @@
> > > #define REFCNT_RING_SIZE (REFCNT_MBUF_NUM * REFCNT_MAX_REF)
> > >
> > > #define MAKE_STRING(x) # x
> > > +#define MBUF_POOL_LOCAL_CACHE_SIZE 32
> > >
> > > static struct rte_mempool *pktmbuf_pool = NULL;
> > >
> > > @@ -405,6 +406,84 @@ test_pktmbuf_pool(void)
> > > return ret;
> > > }
> > >
> > ><snip>
> > > diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
> > > index 17ba791..fabeae2 100644
> > > --- a/lib/librte_mbuf/rte_mbuf.h
> > > +++ b/lib/librte_mbuf/rte_mbuf.h
> > > @@ -825,6 +825,97 @@ static inline void rte_pktmbuf_free(struct rte_mbuf
> > *m)
> > > }
> > >
> > > /**
> > > + * Allocate a bulk of mbufs, initiate refcnt and resets
> > > + *
> > > + * @param pool
> > > + * memory pool to allocate from
> > > + * @param mbufs
> > > + * Array of pointers to mbuf
> > > + * @param count
> > > + * Array size
> > > + */
> > > +static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
> > > + struct rte_mbuf **mbufs,
> > > + unsigned count)
> > > +{
> > > + unsigned idx;
> > > + int rc = 0;
> > > +
> > > + rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
> > > + if (unlikely(rc))
> > > + return rc;
> > > +
> > > + for (idx = 0; idx < count; idx++) {
> > > + RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
> > > + rte_mbuf_refcnt_set(mbufs[idx], 1);
> > > + rte_pktmbuf_reset(mbufs[idx]);
> > > + }
> > > + return rc;
> > > +}
> > > +
> > > +/**
> > > + * Free a bulk of mbufs into its original mempool.
> > > + * This function assumes:
> > > + * - refcnt equals 1
> > > + * - mbufs are direct
> > > + * - all mbufs must belong to the same mempool
> > > + *
> > > + * @param mbufs
> > > + * Array of pointers to mbuf
> > > + * @param count
> > > + * Array size
> > > + */
> > > +static inline void rte_pktmbuf_bulk_free(struct rte_mbuf **mbufs,
> > > + unsigned count)
> > > +{
> > > + unsigned idx;
> > > +
> > > + RTE_MBUF_ASSERT(count > 0);
> > > +
> > > + for (idx = 0; idx < count; idx++) {
> > > + RTE_MBUF_ASSERT(mbufs[idx]->pool == mbufs[0]->pool);
> > > + RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 1);
> > > + rte_mbuf_refcnt_set(mbufs[idx], 0);
> > This is really a misuse of the API. The entire point of reference
> > counting is
> > to know when an mbuf has no more references and can be freed. By forcing
> > all
> > the reference counts to zero here, you allow the refcnt infrastructure to
> > be
> > circumvented, causing memory leaks.
> >
> > I think what you need to do here is enhance the underlying pktmbuf
> > interface
> > such that an rte_mbuf structure has a destructor method association with it
> > which is called when its refcnt reaches zero. That way the
> > rte_pktmbuf_bulk_free function can just decrement the refcnt on each
> > mbuf_structure, and the pool as a whole can be returned when the destructor
> > function discovers that all mbufs in that bulk pool are freed.
> >
> > Neil
> >
> >
