From: "Ananyev, Konstantin" <konstantin.ananyev@intel.com>
To: Hiroyuki MIKITA <h.mikita89@gmail.com>,
"olivier.matz@6wind.com" <olivier.matz@6wind.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH] mbuf: decrease refcnt when detaching
Date: Mon, 16 May 2016 08:49:59 +0000 [thread overview]
Message-ID: <2601191342CEEE43887BDE71AB97725836B5109C@irsmsx105.ger.corp.intel.com> (raw)
In-Reply-To: <CAO_n4D11aSBKpkV8v0S4g84K=QDyUjnqAYhXoZiR+u8BnUrfzQ@mail.gmail.com>
Hi Hiroyuki,
>
> Hi Konstantin,
>
> Now, the attach operation increases refcnt, but the detach does not decrease it.
> I think both operations should affect refcnt or not.
> Which design is intended?
>
> In "6.7. Direct and Indirect Buffers" of Programmer's Guide,
> it is mentioned that "...whenever an indirect buffer is attached to
> the direct buffer,
> the reference counter on the direct buffer is incremented.
> Similarly, whenever the indirect buffer is detached,
> the reference counter on the direct buffer is decremented."
Well, right now the rte_pktmbuf_detach(struct rte_mbuf *m) just restores
the fields of indirect mbufs to the default values, nothing more.
Actual freeing of the mbuf it was attached to is done in the __rte_pktmbuf_prefree_seg().
I suppose the name rte_pktmbuf_detach() is a bit confusing here,
might be, when created, it should be named rte_pktmbuf_restore() or so.
About proposed changes - it would introduce an extra unnecessary read of refcnt (as it is a volatile field).
If we'll decide to go that way, then I think rte_pktmbuf_detach() have to deal with freeing md.
Something like that:
static inline void
rte_pktmbuf_detach(struct rte_mbuf *m)
{
struct rte_mbuf *md = rte_mbuf_from_indirect(m);
/* former rte_pktmbuf_detach */
rte_pktmbuf_restore(m);
if (rte_mbuf_refcnt_update(md, -1) == 0)
__rte_mbuf_raw_free(md);
}
That might be a good thing in terms of API usability and clearness,
but would cause a change in public API behaviour, so I am not sure it is worth it.
Konstantin
>
> Regards,
> Hiroyuki
>
> 2016-05-16 9:05 GMT+09:00 Ananyev, Konstantin <konstantin.ananyev@intel.com>:
> >
> >
> >> -----Original Message-----
> >> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Hiroyuki Mikita
> >> Sent: Sunday, May 15, 2016 4:51 PM
> >> To: olivier.matz@6wind.com
> >> Cc: dev@dpdk.org
> >> Subject: [dpdk-dev] [PATCH] mbuf: decrease refcnt when detaching
> >>
> >> The rte_pktmbuf_detach() function should decrease refcnt on a direct
> >> buffer.
> >
> > Hmm, why is that?
> > What's wrong with the current approach?
> > Konstantin
> >
> >>
> >> Signed-off-by: Hiroyuki Mikita <h.mikita89@gmail.com>
> >> ---
> >> lib/librte_mbuf/rte_mbuf.h | 4 +++-
> >> 1 file changed, 3 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
> >> index 529debb..3b117ca 100644
> >> --- a/lib/librte_mbuf/rte_mbuf.h
> >> +++ b/lib/librte_mbuf/rte_mbuf.h
> >> @@ -1468,9 +1468,11 @@ static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
> >> */
> >> static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
> >> {
> >> + struct rte_mbuf *md = rte_mbuf_from_indirect(m);
> >> struct rte_mempool *mp = m->pool;
> >> uint32_t mbuf_size, buf_len, priv_size;
> >>
> >> + rte_mbuf_refcnt_update(md, -1);
> >> priv_size = rte_pktmbuf_priv_size(mp);
> >> mbuf_size = sizeof(struct rte_mbuf) + priv_size;
> >> buf_len = rte_pktmbuf_data_room_size(mp);
> >> @@ -1498,7 +1500,7 @@ __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
> >> if (RTE_MBUF_INDIRECT(m)) {
> >> struct rte_mbuf *md = rte_mbuf_from_indirect(m);
> >> rte_pktmbuf_detach(m);
> >> - if (rte_mbuf_refcnt_update(md, -1) == 0)
> >> + if (rte_mbuf_refcnt_read(md) == 0)
> >> __rte_mbuf_raw_free(md);
> >> }
> >> return m;
> >> --
> >> 1.9.1
> >
next prev parent reply other threads:[~2016-05-16 8:50 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-05-15 15:50 Hiroyuki Mikita
2016-05-16 0:05 ` Ananyev, Konstantin
2016-05-16 2:46 ` Hiroyuki MIKITA
2016-05-16 8:49 ` Ananyev, Konstantin [this message]
2016-05-16 9:13 ` Thomas Monjalon
2016-05-16 16:24 ` Hiroyuki MIKITA
2016-05-16 8:52 ` Olivier Matz
2016-05-16 16:53 ` [dpdk-dev] [PATCH v2] " Hiroyuki Mikita
2016-05-17 10:58 ` Bruce Richardson
2016-05-17 11:06 ` Ananyev, Konstantin
2016-05-17 12:43 ` Thomas Monjalon
2016-05-17 12:59 ` Ananyev, Konstantin
2016-05-17 13:39 ` Thomas Monjalon
2016-05-17 13:44 ` Ananyev, Konstantin
2016-05-17 14:19 ` Thomas Monjalon
2016-05-17 15:45 ` Ananyev, Konstantin
2016-05-17 16:12 ` Hiroyuki MIKITA
2016-05-17 16:35 ` [dpdk-dev] [PATCH v3] " Hiroyuki Mikita
2016-05-18 11:58 ` Olivier Matz
2016-05-18 14:29 ` Hiroyuki Mikita
2016-05-18 14:41 ` [dpdk-dev] [PATCH v4] " Hiroyuki Mikita
2016-05-18 15:51 ` Olivier Matz
2016-05-19 12:38 ` Thomas Monjalon
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=2601191342CEEE43887BDE71AB97725836B5109C@irsmsx105.ger.corp.intel.com \
--to=konstantin.ananyev@intel.com \
--cc=dev@dpdk.org \
--cc=h.mikita89@gmail.com \
--cc=olivier.matz@6wind.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).