From: Paolo Abeni <pabeni@redhat.com>
To: Kui-Feng Lee <sinquersw@gmail.com>,
Sebastian Andrzej Siewior <bigeasy@linutronix.de>,
netdev@vger.kernel.org, bpf@vger.kernel.org
Cc: "David S. Miller" <davem@davemloft.net>,
Alexei Starovoitov <ast@kernel.org>,
Daniel Borkmann <daniel@iogearbox.net>,
Eric Dumazet <edumazet@google.com>,
Jakub Kicinski <kuba@kernel.org>,
Jesper Dangaard Brouer <hawk@kernel.org>,
John Fastabend <john.fastabend@gmail.com>,
Thomas Gleixner <tglx@linutronix.de>,
Geetha sowjanya <gakula@marvell.com>,
Subbaraya Sundeep <sbhatta@marvell.com>,
Sunil Goutham <sgoutham@marvell.com>,
hariprasad <hkelam@marvell.com>
Subject: Re: [PATCH net v2 3/3] octeontx2-pf: Do xdp_do_flush() after redirects.
Date: Thu, 21 Sep 2023 09:01:16 +0200 [thread overview]
Message-ID: <f5832f155f3daff50b0e7ea94fee84d3a70e4f29.camel@redhat.com> (raw)
In-Reply-To: <aa182e22-e7b9-d8e7-04ea-781fe0fb9103@gmail.com>
On Mon, 2023-09-18 at 10:58 -0700, Kui-Feng Lee wrote:
>
> On 9/18/23 08:36, Sebastian Andrzej Siewior wrote:
> > xdp_do_flush() should be invoked before leaving the NAPI poll function
> > if XDP-redirect has been performed.
> >
> > Invoke xdp_do_flush() before leaving NAPI.
> >
> > Cc: Geetha sowjanya <gakula@marvell.com>
> > Cc: Subbaraya Sundeep <sbhatta@marvell.com>
> > Cc: Sunil Goutham <sgoutham@marvell.com>
> > Cc: hariprasad <hkelam@marvell.com>
> > Fixes: 06059a1a9a4a5 ("octeontx2-pf: Add XDP support to netdev PF")
> > Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> > Acked-by: Geethasowjanya Akula <gakula@marvell.com>
> > ---
> > .../marvell/octeontx2/nic/otx2_txrx.c | 19 +++++++++++++------
> > 1 file changed, 13 insertions(+), 6 deletions(-)
> >
> > diff --git a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c
> > index e77d438489557..53b2a4ef52985 100644
> > --- a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c
> > +++ b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c
> > @@ -29,7 +29,8 @@
> > static bool otx2_xdp_rcv_pkt_handler(struct otx2_nic *pfvf,
> > struct bpf_prog *prog,
> > struct nix_cqe_rx_s *cqe,
> > - struct otx2_cq_queue *cq);
> > + struct otx2_cq_queue *cq,
> > + bool *need_xdp_flush);
> >
> > static int otx2_nix_cq_op_status(struct otx2_nic *pfvf,
> > struct otx2_cq_queue *cq)
> > @@ -337,7 +338,7 @@ static bool otx2_check_rcv_errors(struct otx2_nic *pfvf,
> > static void otx2_rcv_pkt_handler(struct otx2_nic *pfvf,
> > struct napi_struct *napi,
> > struct otx2_cq_queue *cq,
> > - struct nix_cqe_rx_s *cqe)
> > + struct nix_cqe_rx_s *cqe, bool *need_xdp_flush)
> > {
> > struct nix_rx_parse_s *parse = &cqe->parse;
> > struct nix_rx_sg_s *sg = &cqe->sg;
> > @@ -353,7 +354,7 @@ static void otx2_rcv_pkt_handler(struct otx2_nic *pfvf,
> > }
> >
> > if (pfvf->xdp_prog)
> > - if (otx2_xdp_rcv_pkt_handler(pfvf, pfvf->xdp_prog, cqe, cq))
> > + if (otx2_xdp_rcv_pkt_handler(pfvf, pfvf->xdp_prog, cqe, cq, need_xdp_flush))
> > return;
> >
> > skb = napi_get_frags(napi);
> > @@ -388,6 +389,7 @@ static int otx2_rx_napi_handler(struct otx2_nic *pfvf,
> > struct napi_struct *napi,
> > struct otx2_cq_queue *cq, int budget)
> > {
> > + bool need_xdp_flush = false;
> > struct nix_cqe_rx_s *cqe;
> > int processed_cqe = 0;
> >
> > @@ -409,13 +411,15 @@ static int otx2_rx_napi_handler(struct otx2_nic *pfvf,
> > cq->cq_head++;
> > cq->cq_head &= (cq->cqe_cnt - 1);
> >
> > - otx2_rcv_pkt_handler(pfvf, napi, cq, cqe);
> > + otx2_rcv_pkt_handler(pfvf, napi, cq, cqe, &need_xdp_flush);
> >
> > cqe->hdr.cqe_type = NIX_XQE_TYPE_INVALID;
> > cqe->sg.seg_addr = 0x00;
> > processed_cqe++;
> > cq->pend_cqe--;
> > }
> > + if (need_xdp_flush)
> > + xdp_do_flush();
> >
> > /* Free CQEs to HW */
> > otx2_write64(pfvf, NIX_LF_CQ_OP_DOOR,
> > @@ -1354,7 +1358,8 @@ bool otx2_xdp_sq_append_pkt(struct otx2_nic *pfvf, u64 iova, int len, u16 qidx)
> > static bool otx2_xdp_rcv_pkt_handler(struct otx2_nic *pfvf,
> > struct bpf_prog *prog,
> > struct nix_cqe_rx_s *cqe,
> > - struct otx2_cq_queue *cq)
> > + struct otx2_cq_queue *cq,
> > + bool *need_xdp_flush)
> > {
> > unsigned char *hard_start, *data;
> > int qidx = cq->cq_idx;
> > @@ -1391,8 +1396,10 @@ static bool otx2_xdp_rcv_pkt_handler(struct otx2_nic *pfvf,
> >
> > otx2_dma_unmap_page(pfvf, iova, pfvf->rbsize,
> > DMA_FROM_DEVICE);
> > - if (!err)
> > + if (!err) {
> > + *need_xdp_flush = true;
>
> Is it possible to call xdp_do_flush() at the first place (here)?
AFAICT that would kill much/all of the performance benefits of bulk
redirect.
I think the proposed solution is a better one.
Cheers,
Paolo
next prev parent reply other threads:[~2023-09-21 17:49 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-09-18 15:36 [PATCH net v2 0/3] Add missing xdp_do_flush() invocations Sebastian Andrzej Siewior
2023-09-18 15:36 ` [PATCH net v2 1/3] net: ena: Flush XDP packets on error Sebastian Andrzej Siewior
2023-09-18 15:36 ` [PATCH net v2 2/3] bnxt_en: Flush XDP for bnxt_poll_nitroa0()'s NAPI Sebastian Andrzej Siewior
2023-09-18 17:11 ` Michael Chan
2023-09-18 15:36 ` [PATCH net v2 3/3] octeontx2-pf: Do xdp_do_flush() after redirects Sebastian Andrzej Siewior
2023-09-18 17:58 ` Kui-Feng Lee
2023-09-19 6:36 ` Sebastian Andrzej Siewior
2023-09-21 7:01 ` Paolo Abeni [this message]
2023-09-20 7:04 ` [PATCH net v2 0/3] Add missing xdp_do_flush() invocations Jesper Dangaard Brouer
2023-09-20 7:54 ` Sebastian Andrzej Siewior
2023-09-21 7:40 ` patchwork-bot+netdevbpf
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=f5832f155f3daff50b0e7ea94fee84d3a70e4f29.camel@redhat.com \
--to=pabeni@redhat.com \
--cc=ast@kernel.org \
--cc=bigeasy@linutronix.de \
--cc=bpf@vger.kernel.org \
--cc=daniel@iogearbox.net \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=gakula@marvell.com \
--cc=hawk@kernel.org \
--cc=hkelam@marvell.com \
--cc=john.fastabend@gmail.com \
--cc=kuba@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=sbhatta@marvell.com \
--cc=sgoutham@marvell.com \
--cc=sinquersw@gmail.com \
--cc=tglx@linutronix.de \
/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).