From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Zhang, Qi Z" Subject: Re: [PATCH v8 1/3] net/i40e: enable per dev PTYPE mapping table Date: Fri, 7 Apr 2017 02:36:58 +0000 Message-ID: <039ED4275CED7440929022BC67E7061153088D3A@SHSMSX103.ccr.corp.intel.com> References: <20170404015516.7784-1-qi.z.zhang@intel.com> <20170407032505.61691-1-qi.z.zhang@intel.com> <20170407032505.61691-2-qi.z.zhang@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Cc: "Yigit, Ferruh" , "jianbo.liu@linaro.org" , "dev@dpdk.org" , "Wu, Jingjing" , "Zhang, Helin" To: "chaozhu@linux.vnet.ibm.com" Return-path: Received: from mga01.intel.com (mga01.intel.com [192.55.52.88]) by dpdk.org (Postfix) with ESMTP id 4C6D02B87 for ; Fri, 7 Apr 2017 04:37:07 +0200 (CEST) In-Reply-To: <20170407032505.61691-2-qi.z.zhang@intel.com> Content-Language: en-US List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Hi Chao: Would you help to check the PowerPC part. (i40e_rxtx_vec_altivec.c) Thanks Qi > -----Original Message----- > From: Zhang, Qi Z > Sent: Friday, April 7, 2017 11:25 AM > To: Wu, Jingjing ; Zhang, Helin > > Cc: Yigit, Ferruh ; jianbo.liu@linaro.org; > chaozhu@linux.vnet.ibm.com; dev@dpdk.org; Zhang, Qi Z > > Subject: [PATCH v8 1/3] net/i40e: enable per dev PTYPE mapping table >=20 > The mapping from hardware defined packet type to software defined packet > type is static for i40e device, the patch let each ethdev to to have thei= r own > copy of mapping table, this give the possibility that different ethdev ca= n be > set different PTYPE mapping rule which is the requirement to support > following hardware's dynamic PTYPE feature. >=20 > Signed-off-by: Qi Zhang > --- >=20 > v8: >=20 > - Fix compile error for ARM vPMD. >=20 > v7: >=20 > - Add Altivec vPMD support. >=20 > v5: >=20 > - Rebase to dpdk-next-net >=20 > drivers/net/i40e/i40e_ethdev.c | 1 + > drivers/net/i40e/i40e_ethdev.h | 5 +++++ > drivers/net/i40e/i40e_ethdev_vf.c | 2 +- > drivers/net/i40e/i40e_rxtx.c | 30 > ++++++++++++++++++++++-------- > drivers/net/i40e/i40e_rxtx.h | 3 ++- > drivers/net/i40e/i40e_rxtx_vec_altivec.c | 18 ++++++++---------- > drivers/net/i40e/i40e_rxtx_vec_neon.c | 8 +++++--- > drivers/net/i40e/i40e_rxtx_vec_sse.c | 14 ++++++++------ > 8 files changed, 52 insertions(+), 29 deletions(-) >=20 > diff --git a/drivers/net/i40e/i40e_ethdev.c b/drivers/net/i40e/i40e_ethde= v.c > index 6927fde..2b95996 100644 > --- a/drivers/net/i40e/i40e_ethdev.c > +++ b/drivers/net/i40e/i40e_ethdev.c > @@ -1065,6 +1065,7 @@ eth_i40e_dev_init(struct rte_eth_dev *dev) > i40e_set_tx_function(dev); > return 0; > } > + i40e_set_default_ptype_table(dev); > pci_dev =3D I40E_DEV_TO_PCI(dev); > intr_handle =3D &pci_dev->intr_handle; >=20 > diff --git a/drivers/net/i40e/i40e_ethdev.h b/drivers/net/i40e/i40e_ethde= v.h > index 69c6684..a1535b0 100644 > --- a/drivers/net/i40e/i40e_ethdev.h > +++ b/drivers/net/i40e/i40e_ethdev.h > @@ -744,6 +744,8 @@ struct i40e_vf { > uint64_t flags; > }; >=20 > +#define I40E_MAX_PKT_TYPE 256 > + > /* > * Structure to store private data for each PF/VF instance. > */ > @@ -768,6 +770,9 @@ struct i40e_adapter { > struct rte_timecounter systime_tc; > struct rte_timecounter rx_tstamp_tc; > struct rte_timecounter tx_tstamp_tc; > + > + /* ptype mapping table */ > + uint32_t ptype_tbl[I40E_MAX_PKT_TYPE] __rte_cache_min_aligned; > }; >=20 > extern const struct rte_flow_ops i40e_flow_ops; diff --git > a/drivers/net/i40e/i40e_ethdev_vf.c b/drivers/net/i40e/i40e_ethdev_vf.c > index 7e48fea..3db5b6f 100644 > --- a/drivers/net/i40e/i40e_ethdev_vf.c > +++ b/drivers/net/i40e/i40e_ethdev_vf.c > @@ -1477,7 +1477,7 @@ i40evf_dev_init(struct rte_eth_dev *eth_dev) > i40e_set_tx_function(eth_dev); > return 0; > } > - > + i40e_set_default_ptype_table(eth_dev); > rte_eth_copy_pci_info(eth_dev, pci_dev); > eth_dev->data->dev_flags |=3D RTE_ETH_DEV_DETACHABLE; >=20 > diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c = index > ff70c06..e5471b1 100644 > --- a/drivers/net/i40e/i40e_rxtx.c > +++ b/drivers/net/i40e/i40e_rxtx.c > @@ -61,7 +61,6 @@ >=20 > #define DEFAULT_TX_RS_THRESH 32 > #define DEFAULT_TX_FREE_THRESH 32 > -#define I40E_MAX_PKT_TYPE 256 >=20 > #define I40E_TX_MAX_BURST 32 >=20 > @@ -458,6 +457,7 @@ i40e_rx_scan_hw_ring(struct i40e_rx_queue *rxq) > int32_t s[I40E_LOOK_AHEAD], nb_dd; > int32_t i, j, nb_rx =3D 0; > uint64_t pkt_flags; > + uint32_t *ptype_tbl =3D rxq->vsi->adapter->ptype_tbl; >=20 > rxdp =3D &rxq->rx_ring[rxq->rx_tail]; > rxep =3D &rxq->sw_ring[rxq->rx_tail]; > @@ -506,9 +506,9 @@ i40e_rx_scan_hw_ring(struct i40e_rx_queue *rxq) > pkt_flags =3D i40e_rxd_status_to_pkt_flags(qword1); > pkt_flags |=3D i40e_rxd_error_to_pkt_flags(qword1); > mb->packet_type =3D > - i40e_rxd_pkt_type_mapping((uint8_t)((qword1 & > - I40E_RXD_QW1_PTYPE_MASK) >> > - I40E_RXD_QW1_PTYPE_SHIFT)); > + ptype_tbl[(uint8_t)((qword1 & > + I40E_RXD_QW1_PTYPE_MASK) >> > + I40E_RXD_QW1_PTYPE_SHIFT)]; > if (pkt_flags & PKT_RX_RSS_HASH) > mb->hash.rss =3D rte_le_to_cpu_32(\ > rxdp[j].wb.qword0.hi_dword.rss); > @@ -700,6 +700,7 @@ i40e_recv_pkts(void *rx_queue, struct rte_mbuf > **rx_pkts, uint16_t nb_pkts) > uint16_t rx_id, nb_hold; > uint64_t dma_addr; > uint64_t pkt_flags; > + uint32_t *ptype_tbl; >=20 > nb_rx =3D 0; > nb_hold =3D 0; > @@ -707,6 +708,7 @@ i40e_recv_pkts(void *rx_queue, struct rte_mbuf > **rx_pkts, uint16_t nb_pkts) > rx_id =3D rxq->rx_tail; > rx_ring =3D rxq->rx_ring; > sw_ring =3D rxq->sw_ring; > + ptype_tbl =3D rxq->vsi->adapter->ptype_tbl; >=20 > while (nb_rx < nb_pkts) { > rxdp =3D &rx_ring[rx_id]; > @@ -763,8 +765,8 @@ i40e_recv_pkts(void *rx_queue, struct rte_mbuf > **rx_pkts, uint16_t nb_pkts) > pkt_flags =3D i40e_rxd_status_to_pkt_flags(qword1); > pkt_flags |=3D i40e_rxd_error_to_pkt_flags(qword1); > rxm->packet_type =3D > - i40e_rxd_pkt_type_mapping((uint8_t)((qword1 & > - I40E_RXD_QW1_PTYPE_MASK) >> > I40E_RXD_QW1_PTYPE_SHIFT)); > + ptype_tbl[(uint8_t)((qword1 & > + I40E_RXD_QW1_PTYPE_MASK) >> > I40E_RXD_QW1_PTYPE_SHIFT)]; > if (pkt_flags & PKT_RX_RSS_HASH) > rxm->hash.rss =3D > rte_le_to_cpu_32(rxd.wb.qword0.hi_dword.rss); > @@ -818,6 +820,7 @@ i40e_recv_scattered_pkts(void *rx_queue, > uint64_t qword1; > uint64_t dma_addr; > uint64_t pkt_flags; > + uint32_t *ptype_tbl =3D rxq->vsi->adapter->ptype_tbl; >=20 > while (nb_rx < nb_pkts) { > rxdp =3D &rx_ring[rx_id]; > @@ -925,8 +928,8 @@ i40e_recv_scattered_pkts(void *rx_queue, > pkt_flags =3D i40e_rxd_status_to_pkt_flags(qword1); > pkt_flags |=3D i40e_rxd_error_to_pkt_flags(qword1); > first_seg->packet_type =3D > - i40e_rxd_pkt_type_mapping((uint8_t)((qword1 & > - I40E_RXD_QW1_PTYPE_MASK) >> > I40E_RXD_QW1_PTYPE_SHIFT)); > + ptype_tbl[(uint8_t)((qword1 & > + I40E_RXD_QW1_PTYPE_MASK) >> > I40E_RXD_QW1_PTYPE_SHIFT)]; > if (pkt_flags & PKT_RX_RSS_HASH) > first_seg->hash.rss =3D > rte_le_to_cpu_32(rxd.wb.qword0.hi_dword.rss); > @@ -2922,6 +2925,17 @@ i40e_set_tx_function(struct rte_eth_dev *dev) > } > } >=20 > +void __attribute__((cold)) > +i40e_set_default_ptype_table(struct rte_eth_dev *dev) { > + struct i40e_adapter *ad =3D > + I40E_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private); > + int i; > + > + for (i =3D 0; i <=3D I40E_MAX_PKT_TYPE; i++) > + ad->ptype_tbl[i] =3D i40e_get_default_pkt_type(i); } > + > /* Stubs needed for linkage when CONFIG_RTE_I40E_INC_VECTOR is set to > 'n' */ int __attribute__((weak)) > i40e_rx_vec_dev_conf_condition_check(struct rte_eth_dev __rte_unused > *dev) diff --git a/drivers/net/i40e/i40e_rxtx.h b/drivers/net/i40e/i40e_r= xtx.h > index 2fa7f37..20084d6 100644 > --- a/drivers/net/i40e/i40e_rxtx.h > +++ b/drivers/net/i40e/i40e_rxtx.h > @@ -254,13 +254,14 @@ void i40e_set_rx_function(struct rte_eth_dev > *dev); void i40e_set_tx_function_flag(struct rte_eth_dev *dev, > struct i40e_tx_queue *txq); > void i40e_set_tx_function(struct rte_eth_dev *dev); > +void i40e_set_default_ptype_table(struct rte_eth_dev *dev); >=20 > /* For each value it means, datasheet of hardware can tell more details > * > * @note: fix i40e_dev_supported_ptypes_get() if any change here. > */ > static inline uint32_t > -i40e_rxd_pkt_type_mapping(uint8_t ptype) > +i40e_get_default_pkt_type(uint8_t ptype) > { > static const uint32_t type_table[UINT8_MAX + 1] __rte_cache_aligned > =3D { > /* L2 types */ > diff --git a/drivers/net/i40e/i40e_rxtx_vec_altivec.c > b/drivers/net/i40e/i40e_rxtx_vec_altivec.c > index 2f6f70a..5f525ff 100644 > --- a/drivers/net/i40e/i40e_rxtx_vec_altivec.c > +++ b/drivers/net/i40e/i40e_rxtx_vec_altivec.c > @@ -209,7 +209,8 @@ desc_to_olflags_v(vector unsigned long descs[4], > struct rte_mbuf **rx_pkts) > #define PKTLEN_SHIFT 10 >=20 > static inline void > -desc_to_ptype_v(vector unsigned long descs[4], struct rte_mbuf **rx_pkts= ) > +desc_to_ptype_v(vector unsigned long descs[4], struct rte_mbuf **rx_pkts= , > + uint32_t *ptype_tbl) > { > vector unsigned long ptype0 =3D vec_mergel(descs[0], descs[1]); > vector unsigned long ptype1 =3D vec_mergel(descs[2], descs[3]); @@ > -217,14 +218,10 @@ desc_to_ptype_v(vector unsigned long descs[4], struct > rte_mbuf **rx_pkts) > ptype0 =3D vec_sr(ptype0, (vector unsigned long){30, 30}); > ptype1 =3D vec_sr(ptype1, (vector unsigned long){30, 30}); >=20 > - rx_pkts[0]->packet_type =3D i40e_rxd_pkt_type_mapping( > - (*(vector unsigned char *)&ptype0)[0]); > - rx_pkts[1]->packet_type =3D i40e_rxd_pkt_type_mapping( > - (*(vector unsigned char *)&ptype0)[8]); > - rx_pkts[2]->packet_type =3D i40e_rxd_pkt_type_mapping( > - (*(vector unsigned char *)&ptype1)[0]); > - rx_pkts[3]->packet_type =3D i40e_rxd_pkt_type_mapping( > - (*(vector unsigned char *)&ptype1)[8]); > + rx_pkts[0]->packet_type =3D ptype_tbl[(*(vector unsigned char > *)&ptype0)[0])]; > + rx_pkts[1]->packet_type =3D ptype_tbl[(*(vector unsigned char > *)&ptype0)[8])]; > + rx_pkts[2]->packet_type =3D ptype_tbl[(*(vector unsigned char > *)&ptype1)[0])]; > + rx_pkts[3]->packet_type =3D ptype_tbl[(*(vector unsigned char > +*)&ptype1)[8])]; > } >=20 > /* Notice: > @@ -242,6 +239,7 @@ _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, > struct rte_mbuf **rx_pkts, > int pos; > uint64_t var; > vector unsigned char shuf_msk; > + uint32_t *ptype_tbl =3D rxq->vsi->adapter->ptype_tbl; >=20 > vector unsigned short crc_adjust =3D (vector unsigned short){ > 0, 0, /* ignore pkt_type field */ > @@ -466,7 +464,7 @@ _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, > struct rte_mbuf **rx_pkts, > vec_st(pkt_mb1, 0, > (vector unsigned char *)&rx_pkts[pos]->rx_descriptor_fields1 > ); > - desc_to_ptype_v(descs, &rx_pkts[pos]); > + desc_to_ptype_v(descs, &rx_pkts[pos], ptype_tbl); > desc_to_olflags_v(descs, &rx_pkts[pos]); >=20 > /* C.4 calc avaialbe number of desc */ diff --git > a/drivers/net/i40e/i40e_rxtx_vec_neon.c > b/drivers/net/i40e/i40e_rxtx_vec_neon.c > index bd7239b..515931e 100644 > --- a/drivers/net/i40e/i40e_rxtx_vec_neon.c > +++ b/drivers/net/i40e/i40e_rxtx_vec_neon.c > @@ -196,7 +196,8 @@ desc_to_olflags_v(uint64x2_t descs[4], struct > rte_mbuf **rx_pkts) > #define I40E_VPMD_DESC_DD_MASK 0x0001000100010001ULL >=20 > static inline void > -desc_to_ptype_v(uint64x2_t descs[4], struct rte_mbuf **rx_pkts) > +desc_to_ptype_v(uint64x2_t descs[4], struct rte_mbuf **rx_pkts, > + uint32_t *ptype_tbl) > { > int i; > uint8_t ptype; > @@ -205,7 +206,7 @@ desc_to_ptype_v(uint64x2_t descs[4], struct > rte_mbuf **rx_pkts) > for (i =3D 0; i < 4; i++) { > tmp =3D vreinterpretq_u8_u64(vshrq_n_u64(descs[i], 30)); > ptype =3D vgetq_lane_u8(tmp, 8); > - rx_pkts[i]->packet_type =3D i40e_rxd_pkt_type_mapping(ptype); > + rx_pkts[i]->packet_type =3D ptype_tbl[ptype]; > } >=20 > } > @@ -225,6 +226,7 @@ _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, > struct rte_mbuf **rx_pkts, > uint16_t nb_pkts_recd; > int pos; > uint64_t var; > + uint32_t *ptype_tbl =3D rxq->vsi->adapter->ptype_tbl; >=20 > /* mask to shuffle from desc. to mbuf */ > uint8x16_t shuf_msk =3D { > @@ -429,7 +431,7 @@ _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, > struct rte_mbuf **rx_pkts, > pkt_mb2); > vst1q_u8((void *)&rx_pkts[pos]->rx_descriptor_fields1, > pkt_mb1); > - desc_to_ptype_v(descs, &rx_pkts[pos]); > + desc_to_ptype_v(descs, &rx_pkts[pos], ptype_tbl); > /* C.4 calc avaialbe number of desc */ > var =3D __builtin_popcountll(stat & I40E_VPMD_DESC_DD_MASK); > nb_pkts_recd +=3D var; > diff --git a/drivers/net/i40e/i40e_rxtx_vec_sse.c > b/drivers/net/i40e/i40e_rxtx_vec_sse.c > index fdd4a34..335098a 100644 > --- a/drivers/net/i40e/i40e_rxtx_vec_sse.c > +++ b/drivers/net/i40e/i40e_rxtx_vec_sse.c > @@ -215,7 +215,8 @@ desc_to_olflags_v(__m128i descs[4], struct > rte_mbuf **rx_pkts) > #define PKTLEN_SHIFT 10 >=20 > static inline void > -desc_to_ptype_v(__m128i descs[4], struct rte_mbuf **rx_pkts) > +desc_to_ptype_v(__m128i descs[4], struct rte_mbuf **rx_pkts, > + uint32_t *ptype_tbl) > { > __m128i ptype0 =3D _mm_unpackhi_epi64(descs[0], descs[1]); > __m128i ptype1 =3D _mm_unpackhi_epi64(descs[2], descs[3]); @@ > -223,10 +224,10 @@ desc_to_ptype_v(__m128i descs[4], struct rte_mbuf > **rx_pkts) > ptype0 =3D _mm_srli_epi64(ptype0, 30); > ptype1 =3D _mm_srli_epi64(ptype1, 30); >=20 > - rx_pkts[0]->packet_type =3D > i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype0, 0)); > - rx_pkts[1]->packet_type =3D > i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype0, 8)); > - rx_pkts[2]->packet_type =3D > i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype1, 0)); > - rx_pkts[3]->packet_type =3D > i40e_rxd_pkt_type_mapping(_mm_extract_epi8(ptype1, 8)); > + rx_pkts[0]->packet_type =3D ptype_tbl[_mm_extract_epi8(ptype0, 0)]; > + rx_pkts[1]->packet_type =3D ptype_tbl[_mm_extract_epi8(ptype0, 8)]; > + rx_pkts[2]->packet_type =3D ptype_tbl[_mm_extract_epi8(ptype1, 0)]; > + rx_pkts[3]->packet_type =3D ptype_tbl[_mm_extract_epi8(ptype1, 8)]; > } >=20 > /* > @@ -245,6 +246,7 @@ _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, > struct rte_mbuf **rx_pkts, > int pos; > uint64_t var; > __m128i shuf_msk; > + uint32_t *ptype_tbl =3D rxq->vsi->adapter->ptype_tbl; >=20 > __m128i crc_adjust =3D _mm_set_epi16( > 0, 0, 0, /* ignore non-length fields */ > @@ -432,7 +434,7 @@ _recv_raw_pkts_vec(struct i40e_rx_queue *rxq, > struct rte_mbuf **rx_pkts, > pkt_mb2); > _mm_storeu_si128((void *)&rx_pkts[pos]->rx_descriptor_fields1, > pkt_mb1); > - desc_to_ptype_v(descs, &rx_pkts[pos]); > + desc_to_ptype_v(descs, &rx_pkts[pos], ptype_tbl); > /* C.4 calc avaialbe number of desc */ > var =3D __builtin_popcountll(_mm_cvtsi128_si64(staterr)); > nb_pkts_recd +=3D var; > -- > 2.9.3