From: "Michael S. Tsirkin" <mst@redhat.com>
To: John Fastabend <john.fastabend@gmail.com>
Cc: jasowang@redhat.com, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH 2/2] ptr_ring_ll: pop/push multiple objects at once
Date: Tue, 15 Nov 2016 01:06:46 +0200 [thread overview]
Message-ID: <20161115010140-mutt-send-email-mst@kernel.org> (raw)
In-Reply-To: <20161111044432.1547.65342.stgit@john-Precision-Tower-5810>
On Thu, Nov 10, 2016 at 08:44:32PM -0800, John Fastabend wrote:
> Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
This will naturally reduce the cache line bounce
costs, but so will a _many API for ptr-ring,
doing lock-add many-unlock.
the number of atomics also scales better with the lock:
one per push instead of one per queue.
Also, when can qdisc use a _many operation?
> ---
> include/linux/ptr_ring_ll.h | 22 ++++++++++++++++------
> include/linux/skb_array.h | 11 +++++++++--
> net/sched/sch_generic.c | 2 +-
> 3 files changed, 26 insertions(+), 9 deletions(-)
>
> diff --git a/include/linux/ptr_ring_ll.h b/include/linux/ptr_ring_ll.h
> index bcb11f3..5dc25f7 100644
> --- a/include/linux/ptr_ring_ll.h
> +++ b/include/linux/ptr_ring_ll.h
> @@ -45,9 +45,10 @@ struct ptr_ring_ll {
> /* Note: callers invoking this in a loop must use a compiler barrier,
> * for example cpu_relax(). Callers must hold producer_lock.
> */
> -static inline int __ptr_ring_ll_produce(struct ptr_ring_ll *r, void *ptr)
> +static inline int __ptr_ring_ll_produce_many(struct ptr_ring_ll *r,
> + void **ptr, int num)
> {
> - u32 ret, head, tail, next, slots, mask;
> + u32 ret, head, tail, next, slots, mask, i;
>
> do {
> head = READ_ONCE(r->prod_head);
> @@ -55,21 +56,30 @@ static inline int __ptr_ring_ll_produce(struct ptr_ring_ll *r, void *ptr)
> tail = READ_ONCE(r->cons_tail);
>
> slots = mask + tail - head;
> - if (slots < 1)
> + if (slots < num)
> + num = slots;
> +
> + if (unlikely(!num))
> return -ENOMEM;
>
> - next = head + 1;
> + next = head + num;
> ret = cmpxchg(&r->prod_head, head, next);
> } while (ret != head);
>
> - r->queue[head & mask] = ptr;
> + for (i = 0; i < num; i++)
> + r->queue[(head + i) & mask] = ptr[i];
> smp_wmb();
>
> while (r->prod_tail != head)
> cpu_relax();
>
> r->prod_tail = next;
> - return 0;
> + return num;
> +}
> +
> +static inline int __ptr_ring_ll_produce(struct ptr_ring_ll *r, void **ptr)
> +{
> + return __ptr_ring_ll_produce_many(r, ptr, 1);
> }
>
> static inline void *__ptr_ring_ll_consume(struct ptr_ring_ll *r)
> diff --git a/include/linux/skb_array.h b/include/linux/skb_array.h
> index 9b43dfd..de3c700 100644
> --- a/include/linux/skb_array.h
> +++ b/include/linux/skb_array.h
> @@ -48,9 +48,16 @@ static inline bool skb_array_full(struct skb_array *a)
> return ptr_ring_full(&a->ring);
> }
>
> -static inline int skb_array_ll_produce(struct skb_array_ll *a, struct sk_buff *skb)
> +static inline int skb_array_ll_produce_many(struct skb_array_ll *a,
> + struct sk_buff **skb, int num)
> {
> - return __ptr_ring_ll_produce(&a->ring, skb);
> + return __ptr_ring_ll_produce_many(&a->ring, (void **)skb, num);
> +}
> +
> +static inline int skb_array_ll_produce(struct skb_array_ll *a,
> + struct sk_buff **skb)
> +{
> + return __ptr_ring_ll_produce(&a->ring, (void **)skb);
> }
>
> static inline int skb_array_produce(struct skb_array *a, struct sk_buff *skb)
> diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
> index 4648ec8..58f2011 100644
> --- a/net/sched/sch_generic.c
> +++ b/net/sched/sch_generic.c
> @@ -571,7 +571,7 @@ static int pfifo_fast_enqueue(struct sk_buff *skb, struct Qdisc *qdisc,
> struct skb_array_ll *q = band2list(priv, band);
> int err;
>
> - err = skb_array_ll_produce(q, skb);
> + err = skb_array_ll_produce(q, &skb);
>
> if (unlikely(err)) {
> net_warn_ratelimited("drop a packet from fast enqueue\n");
I don't see a pop many operation here.
--
MST
next prev parent reply other threads:[~2016-11-14 23:06 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-11-11 4:43 [RFC PATCH 0/2] illustrate cmpxchg ring for tap/tun and qdisc John Fastabend
2016-11-11 4:44 ` [RFC PATCH 1/2] net: use cmpxchg instead of spinlock in ptr rings John Fastabend
2016-11-14 11:09 ` Jesper Dangaard Brouer
2016-11-14 23:01 ` Michael S. Tsirkin
2016-11-16 4:30 ` John Fastabend
2016-11-11 4:44 ` [RFC PATCH 2/2] ptr_ring_ll: pop/push multiple objects at once John Fastabend
2016-11-14 23:06 ` Michael S. Tsirkin [this message]
2016-11-16 4:42 ` John Fastabend
2016-11-16 5:23 ` Michael S. Tsirkin
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=20161115010140-mutt-send-email-mst@kernel.org \
--to=mst@redhat.com \
--cc=jasowang@redhat.com \
--cc=john.fastabend@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
/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).