From: Olivier Matz <olivier.matz@6wind.com>
To: Maxime Coquelin <maxime.coquelin@redhat.com>,
dev@dpdk.org, yuanhan.liu@linux.intel.com
Cc: konstantin.ananyev@intel.com, sugesh.chandran@intel.com,
bruce.richardson@intel.com, jianfeng.tan@intel.com,
helin.zhang@intel.com, adrien.mazarguil@6wind.com,
stephen@networkplumber.org, dprovan@bivio.net,
xiao.w.wang@intel.com
Subject: Re: [dpdk-dev] [PATCH v2 10/12] virtio: add Tx checksum offload support
Date: Fri, 7 Oct 2016 18:36:35 +0200 [thread overview]
Message-ID: <e413a854-0359-b22a-3f1e-9028ea0c5d15@6wind.com> (raw)
In-Reply-To: <e35e4b43-5052-fafe-febb-a2bf2e92819b@redhat.com>
Hi Maxime,
On 10/07/2016 09:25 AM, Maxime Coquelin wrote:
> Hi Olivier,
>
> On 10/03/2016 11:00 AM, Olivier Matz wrote:
>> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
>> ---
>> drivers/net/virtio/virtio_ethdev.c | 7 +++++
>> drivers/net/virtio/virtio_ethdev.h | 1 +
>> drivers/net/virtio/virtio_rxtx.c | 57
>> +++++++++++++++++++++++++-------------
>> 3 files changed, 45 insertions(+), 20 deletions(-)
>>
>> diff --git a/drivers/net/virtio/virtio_ethdev.c
>> b/drivers/net/virtio/virtio_ethdev.c
>> index 43cb096..55024cd 100644
>> --- a/drivers/net/virtio/virtio_ethdev.c
>> +++ b/drivers/net/virtio/virtio_ethdev.c
>> @@ -1578,6 +1578,13 @@ virtio_dev_info_get(struct rte_eth_dev *dev,
>> struct rte_eth_dev_info *dev_info)
>> dev_info->rx_offload_capa =
>> DEV_RX_OFFLOAD_TCP_CKSUM |
>> DEV_RX_OFFLOAD_UDP_CKSUM;
>> + dev_info->tx_offload_capa = 0;
>> +
>> + if (hw->guest_features & (1ULL << VIRTIO_NET_F_CSUM)) {
>> + dev_info->tx_offload_capa |=
>> + DEV_TX_OFFLOAD_UDP_CKSUM |
>> + DEV_TX_OFFLOAD_TCP_CKSUM;
>> + }
>> }
>>
>> /*
>> diff --git a/drivers/net/virtio/virtio_ethdev.h
>> b/drivers/net/virtio/virtio_ethdev.h
>> index 2fc9218..202aa2e 100644
>> --- a/drivers/net/virtio/virtio_ethdev.h
>> +++ b/drivers/net/virtio/virtio_ethdev.h
>> @@ -62,6 +62,7 @@
>> 1u << VIRTIO_NET_F_CTRL_VQ | \
>> 1u << VIRTIO_NET_F_CTRL_RX | \
>> 1u << VIRTIO_NET_F_CTRL_VLAN | \
>> + 1u << VIRTIO_NET_F_CSUM | \
>> 1u << VIRTIO_NET_F_MRG_RXBUF | \
>> 1ULL << VIRTIO_F_VERSION_1)
>>
>> diff --git a/drivers/net/virtio/virtio_rxtx.c
>> b/drivers/net/virtio/virtio_rxtx.c
>> index eda678a..4ae11e7 100644
>> --- a/drivers/net/virtio/virtio_rxtx.c
>> +++ b/drivers/net/virtio/virtio_rxtx.c
>> @@ -213,13 +213,14 @@ static inline void
>> virtqueue_enqueue_xmit(struct virtnet_tx *txvq, struct rte_mbuf *cookie,
>> uint16_t needed, int use_indirect, int can_push)
>> {
>> + struct virtio_tx_region *txr = txvq->virtio_net_hdr_mz->addr;
>> struct vq_desc_extra *dxp;
>> struct virtqueue *vq = txvq->vq;
>> struct vring_desc *start_dp;
>> uint16_t seg_num = cookie->nb_segs;
>> uint16_t head_idx, idx;
>> uint16_t head_size = vq->hw->vtnet_hdr_size;
>> - unsigned long offs;
>> + struct virtio_net_hdr *hdr;
>>
>> head_idx = vq->vq_desc_head_idx;
>> idx = head_idx;
>> @@ -230,10 +231,9 @@ virtqueue_enqueue_xmit(struct virtnet_tx *txvq,
>> struct rte_mbuf *cookie,
>> start_dp = vq->vq_ring.desc;
>>
>> if (can_push) {
>> - /* put on zero'd transmit header (no offloads) */
>> - void *hdr = rte_pktmbuf_prepend(cookie, head_size);
>> -
>> - memset(hdr, 0, head_size);
>> + /* prepend cannot fail, checked by caller */
>> + hdr = (struct virtio_net_hdr *)
>> + rte_pktmbuf_prepend(cookie, head_size);
>> } else if (use_indirect) {
>> /* setup tx ring slot to point to indirect
>> * descriptor list stored in reserved region.
>> @@ -241,14 +241,11 @@ virtqueue_enqueue_xmit(struct virtnet_tx *txvq,
>> struct rte_mbuf *cookie,
>> * the first slot in indirect ring is already preset
>> * to point to the header in reserved region
>> */
>> - struct virtio_tx_region *txr = txvq->virtio_net_hdr_mz->addr;
>> -
>> - offs = idx * sizeof(struct virtio_tx_region)
>> - + offsetof(struct virtio_tx_region, tx_indir);
>> -
>> - start_dp[idx].addr = txvq->virtio_net_hdr_mem + offs;
>> + start_dp[idx].addr = txvq->virtio_net_hdr_mem +
>> + RTE_PTR_DIFF(&txr[idx].tx_indir, txr);
>> start_dp[idx].len = (seg_num + 1) * sizeof(struct vring_desc);
>> start_dp[idx].flags = VRING_DESC_F_INDIRECT;
>> + hdr = (struct virtio_net_hdr *)&txr[idx].tx_hdr;
>>
>> /* loop below will fill in rest of the indirect elements */
>> start_dp = txr[idx].tx_indir;
>> @@ -257,15 +254,40 @@ virtqueue_enqueue_xmit(struct virtnet_tx *txvq,
>> struct rte_mbuf *cookie,
>> /* setup first tx ring slot to point to header
>> * stored in reserved region.
>> */
>> - offs = idx * sizeof(struct virtio_tx_region)
>> - + offsetof(struct virtio_tx_region, tx_hdr);
>> -
>> - start_dp[idx].addr = txvq->virtio_net_hdr_mem + offs;
>> + start_dp[idx].addr = txvq->virtio_net_hdr_mem +
>> + RTE_PTR_DIFF(&txr[idx].tx_hdr, txr);
>> start_dp[idx].len = vq->hw->vtnet_hdr_size;
>> start_dp[idx].flags = VRING_DESC_F_NEXT;
>> + hdr = (struct virtio_net_hdr *)&txr[idx].tx_hdr;
>> +
>> idx = start_dp[idx].next;
>> }
>>
>> + /* Checksum Offload */
>> + switch (cookie->ol_flags & PKT_TX_L4_MASK) {
>> + case PKT_TX_UDP_CKSUM:
>> + hdr->csum_start = cookie->l2_len + cookie->l3_len;
>> + hdr->csum_offset = 6;
>> + hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
>> + break;
>> +
>> + case PKT_TX_TCP_CKSUM:
>> + hdr->csum_start = cookie->l2_len + cookie->l3_len;
>> + hdr->csum_offset = 16;
>> + hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
>> + break;
>> +
>> + default:
>> + hdr->csum_start = 0;
>> + hdr->csum_offset = 0;
>> + hdr->flags = 0;
>> + break;
>> + }
>> +
>> + hdr->gso_type = 0;
>> + hdr->gso_size = 0;
>> + hdr->hdr_len = 0;
>
> In he case we don't use any offload, have you measured the performance
> regression
> with current code when using a dedicated descriptor for the header?
> I haven't tested you series, but I would think it is more than 15% for
> 64 bytes packets based on my trials to use a single descriptor.
>
> Indeed, without your series, when using a dedicated desc for the header,
> the header is not accessed in the virtio transmit path. It is zeroed at
> init time.
>
> Could we keep the same behaviour when offloading features aren't
> negotiated?
You're right, it could have a performance impact. I'll try to restore
the initial behavior when offload is disabled.
Thanks,
Olivier
next prev parent reply other threads:[~2016-10-07 16:36 UTC|newest]
Thread overview: 97+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-07-21 8:08 [dpdk-dev] [PATCH 00/12] net/virtio: add " Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 01/12] virtio: move device initialization in a function Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 02/12] virtio: setup and start cq in configure callback Olivier Matz
2016-07-21 21:15 ` Stephen Hemminger
2016-07-22 7:54 ` Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 03/12] virtio: reinitialize the device " Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 04/12] mbuf: add function to calculate a checksum Olivier Matz
2016-07-21 10:51 ` Ananyev, Konstantin
2016-07-21 16:26 ` Don Provan
2016-07-21 16:46 ` Olivier Matz
2016-07-22 8:24 ` Olivier Matz
2016-08-29 14:52 ` Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 05/12] mbuf: add new Rx checksum mbuf flags Olivier Matz
2016-07-21 21:22 ` Stephen Hemminger
2016-07-22 8:03 ` Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 06/12] app/testpmd: fix checksum stats in csum engine Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 07/12] mbuf: new flag for LRO Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 08/12] app/testpmd: display lro segment size Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 09/12] virtio: add Rx checksum offload support Olivier Matz
2016-07-27 9:52 ` Wang, Xiao W
2016-07-21 8:08 ` [dpdk-dev] [PATCH 10/12] virtio: add Tx " Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 11/12] virtio: add Lro support Olivier Matz
2016-07-21 8:08 ` [dpdk-dev] [PATCH 12/12] virtio: add Tso support Olivier Matz
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 00/12] net/virtio: add offload support Olivier Matz
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 01/12] virtio: move device initialization in a function Olivier Matz
2016-10-11 12:30 ` Maxime Coquelin
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 02/12] virtio: setup and start cq in configure callback Olivier Matz
2016-10-11 12:47 ` Maxime Coquelin
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 03/12] virtio: reinitialize the device " Olivier Matz
2016-10-11 13:13 ` Maxime Coquelin
2016-10-12 14:41 ` Yuanhan Liu
2016-10-12 16:01 ` Olivier MATZ
2016-10-13 7:54 ` Yuanhan Liu
2016-10-13 13:57 ` Olivier MATZ
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 04/12] net: add function to calculate a checksum in a mbuf Olivier Matz
2016-10-11 13:25 ` Maxime Coquelin
2016-10-11 13:33 ` Olivier MATZ
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 05/12] mbuf: add new Rx checksum mbuf flags Olivier Matz
2016-10-11 13:43 ` Maxime Coquelin
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 06/12] app/testpmd: fix checksum stats in csum engine Olivier Matz
2016-10-11 13:46 ` Maxime Coquelin
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 07/12] mbuf: new flag for LRO Olivier Matz
2016-10-11 13:48 ` Maxime Coquelin
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 08/12] app/testpmd: display lro segment size Olivier Matz
2016-10-11 13:49 ` Maxime Coquelin
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 09/12] virtio: add Rx checksum offload support Olivier Matz
2016-10-03 12:51 ` Maxime Coquelin
2016-10-05 11:56 ` Olivier Matz
2016-10-05 13:27 ` Maxime Coquelin
2016-10-05 13:30 ` Olivier Matz
2016-10-12 13:02 ` Yuanhan Liu
2016-10-12 15:55 ` Olivier MATZ
2016-10-11 14:04 ` Maxime Coquelin
2016-10-11 14:29 ` Olivier MATZ
2016-10-11 14:36 ` Maxime Coquelin
2016-10-11 14:49 ` Olivier MATZ
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 10/12] virtio: add Tx " Olivier Matz
2016-10-07 7:25 ` Maxime Coquelin
2016-10-07 16:36 ` Olivier Matz [this message]
2016-10-13 8:38 ` Yuanhan Liu
2016-10-13 13:58 ` Olivier MATZ
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 11/12] virtio: add Lro support Olivier Matz
2016-10-11 14:21 ` Maxime Coquelin
2016-10-03 9:00 ` [dpdk-dev] [PATCH v2 12/12] virtio: add Tso support Olivier Matz
2016-10-13 8:18 ` Yuanhan Liu
2016-10-13 14:02 ` Olivier MATZ
2016-10-13 14:16 ` Yuanhan Liu
2016-10-13 14:52 ` Olivier MATZ
2016-10-13 15:01 ` Yuanhan Liu
2016-10-13 15:15 ` Olivier MATZ
2016-10-13 15:29 ` Yuanhan Liu
2016-10-13 15:45 ` Olivier Matz
2016-10-13 16:01 ` Yuanhan Liu
2016-10-13 15:04 ` Yuanhan Liu
2016-10-13 23:33 ` Stephen Hemminger
2016-10-18 14:07 ` Olivier Matz
2016-10-11 11:35 ` [dpdk-dev] [PATCH v2 00/12] net/virtio: add offload support Yuanhan Liu
2016-10-11 12:14 ` Olivier MATZ
2016-10-11 15:37 ` Yuanhan Liu
2016-10-13 14:15 ` [dpdk-dev] [PATCH v3 " Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 01/12] net/virtio: move device initialization in a function Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 02/12] net/virtio: setup and start cq in configure callback Olivier Matz
2016-11-02 1:38 ` Yao, Lei A
2016-11-08 14:58 ` Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 03/12] net/virtio: reinitialize the device " Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 04/12] net: add function to calculate a checksum in a mbuf Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 05/12] mbuf: add new Rx checksum mbuf flags Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 06/12] app/testpmd: adapt checksum stats in csum engine Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 07/12] mbuf: new flag for LRO Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 08/12] app/testpmd: display lro segment size Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 09/12] net/virtio: add Rx checksum offload support Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 10/12] net/virtio: add Tx " Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 11/12] net/virtio: add Lro support Olivier Matz
2016-10-13 14:16 ` [dpdk-dev] [PATCH v3 12/12] net/virtio: add Tso support Olivier Matz
2016-10-13 16:05 ` Yuanhan Liu
2016-10-13 18:50 ` Thomas Monjalon
2016-10-13 19:58 ` Olivier Matz
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=e413a854-0359-b22a-3f1e-9028ea0c5d15@6wind.com \
--to=olivier.matz@6wind.com \
--cc=adrien.mazarguil@6wind.com \
--cc=bruce.richardson@intel.com \
--cc=dev@dpdk.org \
--cc=dprovan@bivio.net \
--cc=helin.zhang@intel.com \
--cc=jianfeng.tan@intel.com \
--cc=konstantin.ananyev@intel.com \
--cc=maxime.coquelin@redhat.com \
--cc=stephen@networkplumber.org \
--cc=sugesh.chandran@intel.com \
--cc=xiao.w.wang@intel.com \
--cc=yuanhan.liu@linux.intel.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).