d/numam-dpdk
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

net/mlx5: add no-inline Tx flag

Browse Source 
This patch adds support for dynamic flag that hints transmit
datapath do not copy data to the descriptors. This flag is
useful when data are located in the memory of another (not NIC)
physical device and copying to the host memory is undesirable.

This hint flag is per mbuf for multi-segment packets.

This hint flag might be partially ignored if:

- hardware requires minimal data header to be inline into
  descriptor, it depends on the hardware type and its configuration.
  In this case PMD copies the minimal required number of bytes to
  the descriptor, ignoring the no inline hint flag, the rest of data
  is not copied.

- VLAN tag insertion offload is requested and hardware does not
  support this options. In this case the VLAN tag is inserted by
  software means and at least 18B are copied to descriptor.

Signed-off-by: Viacheslav Ovsiienko <viacheslavo@mellanox.com>
Acked-by: Matan Azrad <matan@mellanox.com>
This commit is contained in:
Viacheslav Ovsiienko 2020-01-29 12:21:07 +00:00 committed by Ferruh Yigit
parent efa79e68c8
commit cacb44a099
1 changed files with 88 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

104
drivers/net/mlx5/mlx5_rxtx.c
View File

@ -130,6 +130,7 @@ uint8_t mlx5_cksum_table[1 << 10] __rte_cache_aligned;
uint8_t mlx5_swp_types_table[1 << 10] __rte_cache_aligned; uint8_t mlx5_swp_types_table[1 << 10] __rte_cache_aligned;
uint64_t rte_net_mlx5_dynf_inline_mask; uint64_t rte_net_mlx5_dynf_inline_mask;
#define PKT_TX_DYNF_NOINLINE rte_net_mlx5_dynf_inline_mask
/** /**
* Build a table to translate Rx completion flags to packet type. * Build a table to translate Rx completion flags to packet type.
@ -2530,21 +2531,30 @@ mlx5_tx_eseg_data(struct mlx5_txq_data *restrict txq,
* Pointer to burst routine local context. * Pointer to burst routine local context.
* @param len * @param len
* Length of data to be copied. * Length of data to be copied.
* @param must
* Length of data to be copied ignoring no inline hint.
* @param olx * @param olx
* Configured Tx offloads mask. It is fully defined at * Configured Tx offloads mask. It is fully defined at
* compile time and may be used for optimization. * compile time and may be used for optimization.
*
* @return
* Number of actual copied data bytes. This is always greater than or
* equal to must parameter and might be lesser than len in no inline
* hint flag is encountered.
*/ */
static __rte_always_inline void static __rte_always_inline unsigned int
mlx5_tx_mseg_memcpy(uint8_t *pdst, mlx5_tx_mseg_memcpy(uint8_t *pdst,
struct mlx5_txq_local *restrict loc, struct mlx5_txq_local *restrict loc,
unsigned int len, unsigned int len,
unsigned int must,
unsigned int olx __rte_unused) unsigned int olx __rte_unused)
{ {
struct rte_mbuf *mbuf; struct rte_mbuf *mbuf;
unsigned int part, dlen; unsigned int part, dlen, copy = 0;
uint8_t *psrc; uint8_t *psrc;
assert(len); assert(len);
assert(must <= len);
do { do {
/* Allow zero length packets, must check first. */ /* Allow zero length packets, must check first. */
dlen = rte_pktmbuf_data_len(loc->mbuf); dlen = rte_pktmbuf_data_len(loc->mbuf);
@ -2557,6 +2567,25 @@ mlx5_tx_mseg_memcpy(uint8_t *pdst,
assert(loc->mbuf_nseg > 1); assert(loc->mbuf_nseg > 1);
assert(loc->mbuf); assert(loc->mbuf);
--loc->mbuf_nseg; --loc->mbuf_nseg;
if (loc->mbuf->ol_flags & PKT_TX_DYNF_NOINLINE) {
unsigned int diff;
if (copy >= must) {
/*
* We already copied the minimal
* requested amount of data.
*/
return copy;
}
diff = must - copy;
if (diff <= rte_pktmbuf_data_len(loc->mbuf)) {
/*
* Copy only the minimal required
* part of the data buffer.
*/
len = diff;
}
}
continue; continue;
} }
dlen -= loc->mbuf_off; dlen -= loc->mbuf_off;
@ -2564,6 +2593,7 @@ mlx5_tx_mseg_memcpy(uint8_t *pdst,
loc->mbuf_off); loc->mbuf_off);
part = RTE_MIN(len, dlen); part = RTE_MIN(len, dlen);
rte_memcpy(pdst, psrc, part); rte_memcpy(pdst, psrc, part);
copy += part;
loc->mbuf_off += part; loc->mbuf_off += part;
len -= part; len -= part;
if (!len) { if (!len) {
@ -2577,7 +2607,7 @@ mlx5_tx_mseg_memcpy(uint8_t *pdst,
assert(loc->mbuf_nseg >= 1); assert(loc->mbuf_nseg >= 1);
--loc->mbuf_nseg; --loc->mbuf_nseg;
} }
return; return copy;
} }
pdst += part; pdst += part;
} while (true); } while (true);
@ -2622,7 +2652,7 @@ mlx5_tx_eseg_mdat(struct mlx5_txq_data *restrict txq,
struct mlx5_wqe_eseg *restrict es = &wqe->eseg; struct mlx5_wqe_eseg *restrict es = &wqe->eseg;
uint32_t csum; uint32_t csum;
uint8_t *pdst; uint8_t *pdst;
unsigned int part; unsigned int part, tlen = 0;
/* /*
* Calculate and set check sum flags first, uint32_t field * Calculate and set check sum flags first, uint32_t field
@ -2655,17 +2685,18 @@ mlx5_tx_eseg_mdat(struct mlx5_txq_data *restrict txq,
2 * RTE_ETHER_ADDR_LEN), 2 * RTE_ETHER_ADDR_LEN),
"invalid Ethernet Segment data size"); "invalid Ethernet Segment data size");
assert(inlen >= MLX5_ESEG_MIN_INLINE_SIZE); assert(inlen >= MLX5_ESEG_MIN_INLINE_SIZE);
es->inline_hdr_sz = rte_cpu_to_be_16(inlen);
pdst = (uint8_t *)&es->inline_data; pdst = (uint8_t *)&es->inline_data;
if (MLX5_TXOFF_CONFIG(VLAN) && vlan) { if (MLX5_TXOFF_CONFIG(VLAN) && vlan) {
/* Implement VLAN tag insertion as part inline data. */ /* Implement VLAN tag insertion as part inline data. */
mlx5_tx_mseg_memcpy(pdst, loc, 2 * RTE_ETHER_ADDR_LEN, olx); mlx5_tx_mseg_memcpy(pdst, loc,
2 * RTE_ETHER_ADDR_LEN,
2 * RTE_ETHER_ADDR_LEN, olx);
pdst += 2 * RTE_ETHER_ADDR_LEN; pdst += 2 * RTE_ETHER_ADDR_LEN;
*(unaligned_uint32_t *)pdst = rte_cpu_to_be_32 *(unaligned_uint32_t *)pdst = rte_cpu_to_be_32
((RTE_ETHER_TYPE_VLAN << 16) | ((RTE_ETHER_TYPE_VLAN << 16) |
loc->mbuf->vlan_tci); loc->mbuf->vlan_tci);
pdst += sizeof(struct rte_vlan_hdr); pdst += sizeof(struct rte_vlan_hdr);
inlen -= 2 * RTE_ETHER_ADDR_LEN + sizeof(struct rte_vlan_hdr); tlen += 2 * RTE_ETHER_ADDR_LEN + sizeof(struct rte_vlan_hdr);
} }
assert(pdst < (uint8_t *)txq->wqes_end); assert(pdst < (uint8_t *)txq->wqes_end);
/* /*
@ -2673,18 +2704,26 @@ mlx5_tx_eseg_mdat(struct mlx5_txq_data *restrict txq,
* Here we should be aware of WQE ring buffer wraparound only. * Here we should be aware of WQE ring buffer wraparound only.
*/ */
part = (uint8_t *)txq->wqes_end - pdst; part = (uint8_t *)txq->wqes_end - pdst;
part = RTE_MIN(part, inlen); part = RTE_MIN(part, inlen - tlen);
assert(part); assert(part);
do { do {
mlx5_tx_mseg_memcpy(pdst, loc, part, olx); unsigned int copy;
inlen -= part;
if (likely(!inlen)) { /*
pdst += part; * Copying may be interrupted inside the routine
* if run into no inline hint flag.
*/
copy = tlen >= txq->inlen_mode ? 0 : (txq->inlen_mode - tlen);
copy = mlx5_tx_mseg_memcpy(pdst, loc, part, copy, olx);
tlen += copy;
if (likely(inlen <= tlen) || copy < part) {
es->inline_hdr_sz = rte_cpu_to_be_16(tlen);
pdst += copy;
pdst = RTE_PTR_ALIGN(pdst, MLX5_WSEG_SIZE); pdst = RTE_PTR_ALIGN(pdst, MLX5_WSEG_SIZE);
return (struct mlx5_wqe_dseg *)pdst; return (struct mlx5_wqe_dseg *)pdst;
} }
pdst = (uint8_t *)txq->wqes; pdst = (uint8_t *)txq->wqes;
part = inlen; part = inlen - tlen;
} while (true); } while (true);
} }
@ -3283,7 +3322,8 @@ mlx5_tx_packet_multi_inline(struct mlx5_txq_data *restrict txq,
if (inlen <= MLX5_ESEG_MIN_INLINE_SIZE) if (inlen <= MLX5_ESEG_MIN_INLINE_SIZE)
return MLX5_TXCMP_CODE_ERROR; return MLX5_TXCMP_CODE_ERROR;
assert(txq->inlen_send >= MLX5_ESEG_MIN_INLINE_SIZE); assert(txq->inlen_send >= MLX5_ESEG_MIN_INLINE_SIZE);
if (inlen > txq->inlen_send) { if (inlen > txq->inlen_send ||
loc->mbuf->ol_flags & PKT_TX_DYNF_NOINLINE) {
struct rte_mbuf *mbuf; struct rte_mbuf *mbuf;
unsigned int nxlen; unsigned int nxlen;
uintptr_t start; uintptr_t start;
@ -3298,7 +3338,8 @@ mlx5_tx_packet_multi_inline(struct mlx5_txq_data *restrict txq,
assert(txq->inlen_mode <= txq->inlen_send); assert(txq->inlen_mode <= txq->inlen_send);
inlen = txq->inlen_mode; inlen = txq->inlen_mode;
} else { } else {
if (!vlan || txq->vlan_en) { if (loc->mbuf->ol_flags & PKT_TX_DYNF_NOINLINE ||
!vlan || txq->vlan_en) {
/* /*
* VLAN insertion will be done inside by HW. * VLAN insertion will be done inside by HW.
* It is not utmost effective - VLAN flag is * It is not utmost effective - VLAN flag is
@ -4109,7 +4150,8 @@ mlx5_tx_burst_empw_inline(struct mlx5_txq_data *restrict txq,
return MLX5_TXCMP_CODE_ERROR; return MLX5_TXCMP_CODE_ERROR;
} }
/* Inline or not inline - that's the Question. */ /* Inline or not inline - that's the Question. */
if (dlen > txq->inlen_empw) if (dlen > txq->inlen_empw ||
loc->mbuf->ol_flags & PKT_TX_DYNF_NOINLINE)
goto pointer_empw; goto pointer_empw;
/* Inline entire packet, optional VLAN insertion. */ /* Inline entire packet, optional VLAN insertion. */
tlen = sizeof(dseg->bcount) + dlen; tlen = sizeof(dseg->bcount) + dlen;
@ -4305,6 +4347,33 @@ mlx5_tx_burst_single_send(struct mlx5_txq_data *restrict txq,
/* Check against minimal length. */ /* Check against minimal length. */
if (inlen <= MLX5_ESEG_MIN_INLINE_SIZE) if (inlen <= MLX5_ESEG_MIN_INLINE_SIZE)
return MLX5_TXCMP_CODE_ERROR; return MLX5_TXCMP_CODE_ERROR;
if (loc->mbuf->ol_flags &
PKT_TX_DYNF_NOINLINE) {
/*
* The hint flag not to inline packet
* data is set. Check whether we can
* follow the hint.
*/
if ((!MLX5_TXOFF_CONFIG(EMPW) &&
txq->inlen_mode) ||
(MLX5_TXOFF_CONFIG(MPW) &&
txq->inlen_mode)) {
/*
* The hardware requires the
* minimal inline data header.
*/
goto single_min_inline;
}
if (MLX5_TXOFF_CONFIG(VLAN) &&
vlan && !txq->vlan_en) {
/*
* We must insert VLAN tag
* by software means.
*/
goto single_part_inline;
}
goto single_no_inline;
}
/* /*
* Completely inlined packet data WQE: * Completely inlined packet data WQE:
* - Control Segment, SEND opcode * - Control Segment, SEND opcode
@ -4354,6 +4423,7 @@ mlx5_tx_burst_single_send(struct mlx5_txq_data *restrict txq,
* We should check the free space in * We should check the free space in
* WQE ring buffer to inline partially. * WQE ring buffer to inline partially.
*/ */
single_min_inline:
assert(txq->inlen_send >= txq->inlen_mode); assert(txq->inlen_send >= txq->inlen_mode);
assert(inlen > txq->inlen_mode); assert(inlen > txq->inlen_mode);
assert(txq->inlen_mode >= assert(txq->inlen_mode >=
@ -4421,6 +4491,7 @@ mlx5_tx_burst_single_send(struct mlx5_txq_data *restrict txq,
* We also get here if VLAN insertion is not * We also get here if VLAN insertion is not
* supported by HW, the inline is enabled. * supported by HW, the inline is enabled.
*/ */
single_part_inline:
wqe = txq->wqes + (txq->wqe_ci & txq->wqe_m); wqe = txq->wqes + (txq->wqe_ci & txq->wqe_m);
loc->wqe_last = wqe; loc->wqe_last = wqe;
mlx5_tx_cseg_init(txq, loc, wqe, 4, mlx5_tx_cseg_init(txq, loc, wqe, 4,
@ -4461,6 +4532,7 @@ mlx5_tx_burst_single_send(struct mlx5_txq_data *restrict txq,
* - Ethernet Segment, optional VLAN, no inline * - Ethernet Segment, optional VLAN, no inline
* - Data Segment, pointer type * - Data Segment, pointer type
*/ */
single_no_inline:
wqe = txq->wqes + (txq->wqe_ci & txq->wqe_m); wqe = txq->wqes + (txq->wqe_ci & txq->wqe_m);
loc->wqe_last = wqe; loc->wqe_last = wqe;
mlx5_tx_cseg_init(txq, loc, wqe, 3, mlx5_tx_cseg_init(txq, loc, wqe, 3,