net/iavf: fix VLAN tag extraction handling in AVX512 path

Message ID 20210422024830.10617-1-leyi.rong@intel.com (mailing list archive)
State Accepted, archived
Delegated to: Qi Zhang
Headers
Series net/iavf: fix VLAN tag extraction handling in AVX512 path |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation success Compilation OK
ci/github-robot success github build: passed
ci/intel-Testing success Testing PASS
ci/iol-testing success Testing PASS
ci/iol-intel-Performance success Performance Testing PASS
ci/iol-abi-testing success Testing PASS
ci/iol-mellanox-Performance success Performance Testing PASS

Commit Message

Leyi Rong April 22, 2021, 2:48 a.m. UTC
  The new VIRTCHNL_VF_OFFLOAD_VLAN_V2 capability added support that allows
the PF to set the location of the RX VLAN tag for stripping offloads.

So the VF needs to extract the VLAN tag according to the location flags.

This patch is the fix for AVX512 path, as AVX2 is already fixed.

Fixes: 9c9aa0040344 ("net/iavf: add offload path for Rx AVX512 flex descriptor")

Signed-off-by: Leyi Rong <leyi.rong@intel.com>
---
 drivers/net/iavf/iavf_rxtx_vec_avx512.c | 215 +++++++++++++++++++-----
 1 file changed, 174 insertions(+), 41 deletions(-)
  

Comments

Sun, QinX April 27, 2021, 7:11 a.m. UTC | #1
> -----Original Message-----
> From: dev <dev-bounces@dpdk.org> On Behalf Of Leyi Rong
> Sent: Thursday, April 22, 2021 10:49 AM
> To: Zhang, Qi Z <qi.z.zhang@intel.com>; Lu, Wenzhuo
> <wenzhuo.lu@intel.com>
> Cc: dev@dpdk.org; Rong, Leyi <leyi.rong@intel.com>
> Subject: [dpdk-dev] [PATCH] net/iavf: fix VLAN tag extraction handling in
> AVX512 path

Tested-by: Qin Sun <qinx.sun@intel.com>
  
Qi Zhang April 28, 2021, 1:58 a.m. UTC | #2
> -----Original Message-----
> From: Sun, QinX <qinx.sun@intel.com>
> Sent: Tuesday, April 27, 2021 3:11 PM
> To: Rong, Leyi <leyi.rong@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>; Lu,
> Wenzhuo <wenzhuo.lu@intel.com>
> Cc: dev@dpdk.org; Rong, Leyi <leyi.rong@intel.com>
> Subject: RE: [dpdk-dev] [PATCH] net/iavf: fix VLAN tag extraction handling in
> AVX512 path
> 
> > -----Original Message-----
> > From: dev <dev-bounces@dpdk.org> On Behalf Of Leyi Rong
> > Sent: Thursday, April 22, 2021 10:49 AM
> > To: Zhang, Qi Z <qi.z.zhang@intel.com>; Lu, Wenzhuo
> > <wenzhuo.lu@intel.com>
> > Cc: dev@dpdk.org; Rong, Leyi <leyi.rong@intel.com>
> > Subject: [dpdk-dev] [PATCH] net/iavf: fix VLAN tag extraction handling
> > in
> > AVX512 path
> 
> Tested-by: Qin Sun <qinx.sun@intel.com>

Please keep the original commit log and signed-off.
  
Sun, QinX April 29, 2021, 6:45 a.m. UTC | #3
> -----Original Message-----
> From: dev <dev-bounces@dpdk.org> On Behalf Of Leyi Rong
> Sent: Thursday, April 22, 2021 10:49 AM
> To: Zhang, Qi Z <qi.z.zhang@intel.com>; Lu, Wenzhuo
> <wenzhuo.lu@intel.com>
> Cc: dev@dpdk.org; Rong, Leyi <leyi.rong@intel.com>
> Subject: [dpdk-dev] [PATCH] net/iavf: fix VLAN tag extraction handling in
> AVX512 path
> 
> The new VIRTCHNL_VF_OFFLOAD_VLAN_V2 capability added support that
> allows the PF to set the location of the RX VLAN tag for stripping offloads.
> 
> So the VF needs to extract the VLAN tag according to the location flags.
> 
> This patch is the fix for AVX512 path, as AVX2 is already fixed.
> 
> Fixes: 9c9aa0040344 ("net/iavf: add offload path for Rx AVX512 flex
> descriptor")
> 
> Signed-off-by: Leyi Rong <leyi.rong@intel.com>

Tested-by: Qin Sun  <qinx.sun@intel.com>
  
Qi Zhang April 30, 2021, 2:07 p.m. UTC | #4
> -----Original Message-----
> From: Sun, QinX <qinx.sun@intel.com>
> Sent: Thursday, April 29, 2021 2:45 PM
> To: Rong, Leyi <leyi.rong@intel.com>; Zhang, Qi Z <qi.z.zhang@intel.com>; Lu,
> Wenzhuo <wenzhuo.lu@intel.com>
> Cc: dev@dpdk.org; Rong, Leyi <leyi.rong@intel.com>
> Subject: RE: [dpdk-dev] [PATCH] net/iavf: fix VLAN tag extraction handling in
> AVX512 path
> 
> > -----Original Message-----
> > From: dev <dev-bounces@dpdk.org> On Behalf Of Leyi Rong
> > Sent: Thursday, April 22, 2021 10:49 AM
> > To: Zhang, Qi Z <qi.z.zhang@intel.com>; Lu, Wenzhuo
> > <wenzhuo.lu@intel.com>
> > Cc: dev@dpdk.org; Rong, Leyi <leyi.rong@intel.com>
> > Subject: [dpdk-dev] [PATCH] net/iavf: fix VLAN tag extraction handling
> > in
> > AVX512 path
> >
> > The new VIRTCHNL_VF_OFFLOAD_VLAN_V2 capability added support that
> > allows the PF to set the location of the RX VLAN tag for stripping offloads.
> >
> > So the VF needs to extract the VLAN tag according to the location flags.
> >
> > This patch is the fix for AVX512 path, as AVX2 is already fixed.
> >
> > Fixes: 9c9aa0040344 ("net/iavf: add offload path for Rx AVX512 flex
> > descriptor")
> >
> > Signed-off-by: Leyi Rong <leyi.rong@intel.com>
> 
> Tested-by: Qin Sun  <qinx.sun@intel.com>

Applied to dpdk-next-net-intel.

Thanks
Qi
  

Patch

diff --git a/drivers/net/iavf/iavf_rxtx_vec_avx512.c b/drivers/net/iavf/iavf_rxtx_vec_avx512.c
index 7faa1f44aa..d99de2a8bc 100644
--- a/drivers/net/iavf/iavf_rxtx_vec_avx512.c
+++ b/drivers/net/iavf/iavf_rxtx_vec_avx512.c
@@ -952,6 +952,7 @@  _iavf_recv_raw_pkts_vec_avx512_flex_rxd(struct iavf_rx_queue *rxq,
 
 		/* merge flags */
 		__m256i mbuf_flags = _mm256_set1_epi32(0);
+		__m256i vlan_flags = _mm256_setzero_si256();
 
 		if (offload) {
 #if defined(IAVF_RX_CSUM_OFFLOAD) || defined(IAVF_RX_VLAN_OFFLOAD) || defined(IAVF_RX_RSS_OFFLOAD)
@@ -1010,19 +1011,33 @@  _iavf_recv_raw_pkts_vec_avx512_flex_rxd(struct iavf_rx_queue *rxq,
 			 * If RSS(bit12)/VLAN(bit13) are set,
 			 * shuffle moves appropriate flags in place.
 			 */
-			const __m256i rss_vlan_flags_shuf = _mm256_set_epi8(0, 0, 0, 0,
-					0, 0, 0, 0,
-					0, 0, 0, 0,
-					PKT_RX_RSS_HASH | PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED,
-					PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED,
-					PKT_RX_RSS_HASH, 0,
-					/* end up 128-bits */
-					0, 0, 0, 0,
-					0, 0, 0, 0,
-					0, 0, 0, 0,
-					PKT_RX_RSS_HASH | PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED,
-					PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED,
-					PKT_RX_RSS_HASH, 0);
+			const __m256i rss_flags_shuf = _mm256_set_epi8
+					(0, 0, 0, 0,
+					 0, 0, 0, 0,
+					 0, 0, 0, 0,
+					 PKT_RX_RSS_HASH, 0,
+					 PKT_RX_RSS_HASH, 0,
+					 /* end up 128-bits */
+					 0, 0, 0, 0,
+					 0, 0, 0, 0,
+					 0, 0, 0, 0,
+					 PKT_RX_RSS_HASH, 0,
+					 PKT_RX_RSS_HASH, 0);
+
+			const __m256i vlan_flags_shuf = _mm256_set_epi8
+					(0, 0, 0, 0,
+					 0, 0, 0, 0,
+					 0, 0, 0, 0,
+					 PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED,
+					 PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED,
+					 0, 0,
+					 /* end up 128-bits */
+					 0, 0, 0, 0,
+					 0, 0, 0, 0,
+					 0, 0, 0, 0,
+					 PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED,
+					 PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED,
+					 0, 0);
 #endif
 
 #if defined(IAVF_RX_CSUM_OFFLOAD) || defined(IAVF_RX_VLAN_OFFLOAD) || defined(IAVF_RX_RSS_OFFLOAD)
@@ -1044,9 +1059,18 @@  _iavf_recv_raw_pkts_vec_avx512_flex_rxd(struct iavf_rx_queue *rxq,
 			/* set rss and vlan flags */
 			const __m256i rss_vlan_flag_bits =
 				_mm256_srli_epi32(flag_bits, 12);
-			const __m256i rss_vlan_flags =
-				_mm256_shuffle_epi8(rss_vlan_flags_shuf,
+			const __m256i rss_flags =
+				_mm256_shuffle_epi8(rss_flags_shuf,
 						    rss_vlan_flag_bits);
+
+			if (rxq->rx_flags == IAVF_RX_FLAGS_VLAN_TAG_LOC_L2TAG1)
+				vlan_flags =
+					_mm256_shuffle_epi8(vlan_flags_shuf,
+							    rss_vlan_flag_bits);
+
+			const __m256i rss_vlan_flags =
+				_mm256_or_si256(rss_flags, vlan_flags);
+
 #endif
 
 #ifdef IAVF_RX_CSUM_OFFLOAD
@@ -1114,7 +1138,8 @@  _iavf_recv_raw_pkts_vec_avx512_flex_rxd(struct iavf_rx_queue *rxq,
 			 * will cause performance drop to get into this context.
 			 */
 			if (rxq->vsi->adapter->eth_dev->data->dev_conf.rxmode.offloads &
-			    DEV_RX_OFFLOAD_RSS_HASH) {
+			    DEV_RX_OFFLOAD_RSS_HASH ||
+			    rxq->rx_flags & IAVF_RX_FLAGS_VLAN_TAG_LOC_L2TAG2_2) {
 				/* load bottom half of every 32B desc */
 				const __m128i raw_desc_bh7 =
 					_mm_load_si128
@@ -1165,36 +1190,144 @@  _iavf_recv_raw_pkts_vec_avx512_flex_rxd(struct iavf_rx_queue *rxq,
 						(_mm256_castsi128_si256(raw_desc_bh0),
 						 raw_desc_bh1, 1);
 
-				/**
-				 * to shift the 32b RSS hash value to the
-				 * highest 32b of each 128b before mask
-				 */
-				__m256i rss_hash6_7 =
-					_mm256_slli_epi64(raw_desc_bh6_7, 32);
-				__m256i rss_hash4_5 =
-					_mm256_slli_epi64(raw_desc_bh4_5, 32);
-				__m256i rss_hash2_3 =
-					_mm256_slli_epi64(raw_desc_bh2_3, 32);
-				__m256i rss_hash0_1 =
-					_mm256_slli_epi64(raw_desc_bh0_1, 32);
-
-				__m256i rss_hash_msk =
-					_mm256_set_epi32(0xFFFFFFFF, 0, 0, 0,
-							 0xFFFFFFFF, 0, 0, 0);
-
-				rss_hash6_7 = _mm256_and_si256
+				if (rxq->vsi->adapter->eth_dev->data->dev_conf.rxmode.offloads &
+						DEV_RX_OFFLOAD_RSS_HASH) {
+					/**
+					 * to shift the 32b RSS hash value to the
+					 * highest 32b of each 128b before mask
+					 */
+					__m256i rss_hash6_7 =
+						_mm256_slli_epi64
+						(raw_desc_bh6_7, 32);
+					__m256i rss_hash4_5 =
+						_mm256_slli_epi64
+						(raw_desc_bh4_5, 32);
+					__m256i rss_hash2_3 =
+						_mm256_slli_epi64
+						(raw_desc_bh2_3, 32);
+					__m256i rss_hash0_1 =
+						_mm256_slli_epi64
+						(raw_desc_bh0_1, 32);
+
+					const __m256i rss_hash_msk =
+						_mm256_set_epi32
+						(0xFFFFFFFF, 0, 0, 0,
+						 0xFFFFFFFF, 0, 0, 0);
+
+					rss_hash6_7 = _mm256_and_si256
 						(rss_hash6_7, rss_hash_msk);
-				rss_hash4_5 = _mm256_and_si256
+					rss_hash4_5 = _mm256_and_si256
 						(rss_hash4_5, rss_hash_msk);
-				rss_hash2_3 = _mm256_and_si256
+					rss_hash2_3 = _mm256_and_si256
 						(rss_hash2_3, rss_hash_msk);
-				rss_hash0_1 = _mm256_and_si256
+					rss_hash0_1 = _mm256_and_si256
 						(rss_hash0_1, rss_hash_msk);
 
-				mb6_7 = _mm256_or_si256(mb6_7, rss_hash6_7);
-				mb4_5 = _mm256_or_si256(mb4_5, rss_hash4_5);
-				mb2_3 = _mm256_or_si256(mb2_3, rss_hash2_3);
-				mb0_1 = _mm256_or_si256(mb0_1, rss_hash0_1);
+					mb6_7 = _mm256_or_si256
+						(mb6_7, rss_hash6_7);
+					mb4_5 = _mm256_or_si256
+						(mb4_5, rss_hash4_5);
+					mb2_3 = _mm256_or_si256
+						(mb2_3, rss_hash2_3);
+					mb0_1 = _mm256_or_si256
+						(mb0_1, rss_hash0_1);
+				}
+
+				if (rxq->rx_flags & IAVF_RX_FLAGS_VLAN_TAG_LOC_L2TAG2_2) {
+					/* merge the status/error-1 bits into one register */
+					const __m256i status1_4_7 =
+						_mm256_unpacklo_epi32
+						(raw_desc_bh6_7,
+						 raw_desc_bh4_5);
+					const __m256i status1_0_3 =
+						_mm256_unpacklo_epi32
+						(raw_desc_bh2_3,
+						 raw_desc_bh0_1);
+
+					const __m256i status1_0_7 =
+						_mm256_unpacklo_epi64
+						(status1_4_7, status1_0_3);
+
+					const __m256i l2tag2p_flag_mask =
+						_mm256_set1_epi32
+						(1 << IAVF_RX_FLEX_DESC_STATUS1_L2TAG2P_S);
+
+					__m256i l2tag2p_flag_bits =
+						_mm256_and_si256
+						(status1_0_7,
+						 l2tag2p_flag_mask);
+
+					l2tag2p_flag_bits =
+						_mm256_srli_epi32
+						(l2tag2p_flag_bits,
+						 IAVF_RX_FLEX_DESC_STATUS1_L2TAG2P_S);
+
+					const __m256i l2tag2_flags_shuf =
+						_mm256_set_epi8
+							(0, 0, 0, 0,
+							 0, 0, 0, 0,
+							 0, 0, 0, 0,
+							 0, 0, 0, 0,
+							 /* end up 128-bits */
+							 0, 0, 0, 0,
+							 0, 0, 0, 0,
+							 0, 0, 0, 0,
+							 0, 0,
+							 PKT_RX_VLAN |
+							 PKT_RX_VLAN_STRIPPED,
+							 0);
+
+					vlan_flags =
+						_mm256_shuffle_epi8
+							(l2tag2_flags_shuf,
+							 l2tag2p_flag_bits);
+
+					/* merge with vlan_flags */
+					mbuf_flags = _mm256_or_si256
+							(mbuf_flags,
+							 vlan_flags);
+
+					/* L2TAG2_2 */
+					__m256i vlan_tci6_7 =
+						_mm256_slli_si256
+							(raw_desc_bh6_7, 4);
+					__m256i vlan_tci4_5 =
+						_mm256_slli_si256
+							(raw_desc_bh4_5, 4);
+					__m256i vlan_tci2_3 =
+						_mm256_slli_si256
+							(raw_desc_bh2_3, 4);
+					__m256i vlan_tci0_1 =
+						_mm256_slli_si256
+							(raw_desc_bh0_1, 4);
+
+					const __m256i vlan_tci_msk =
+						_mm256_set_epi32
+						(0, 0xFFFF0000, 0, 0,
+						 0, 0xFFFF0000, 0, 0);
+
+					vlan_tci6_7 = _mm256_and_si256
+							(vlan_tci6_7,
+							 vlan_tci_msk);
+					vlan_tci4_5 = _mm256_and_si256
+							(vlan_tci4_5,
+							 vlan_tci_msk);
+					vlan_tci2_3 = _mm256_and_si256
+							(vlan_tci2_3,
+							 vlan_tci_msk);
+					vlan_tci0_1 = _mm256_and_si256
+							(vlan_tci0_1,
+							 vlan_tci_msk);
+
+					mb6_7 = _mm256_or_si256
+							(mb6_7, vlan_tci6_7);
+					mb4_5 = _mm256_or_si256
+							(mb4_5, vlan_tci4_5);
+					mb2_3 = _mm256_or_si256
+							(mb2_3, vlan_tci2_3);
+					mb0_1 = _mm256_or_si256
+							(mb0_1, vlan_tci0_1);
+				}
 			} /* if() on RSS hash parsing */
 #endif
 		}