1
0
mirror of https://github.com/corundum/corundum.git synced 2025-01-30 08:32:52 +08:00

Update receive handling to allocate pages instead of skbs

This commit is contained in:
Alex Forencich 2020-03-10 23:06:54 -07:00
parent 8536b7d2b7
commit 65ead3a064
3 changed files with 58 additions and 39 deletions

View File

@ -121,9 +121,11 @@ struct mqnic_tx_info {
}; };
struct mqnic_rx_info { struct mqnic_rx_info {
struct sk_buff *skb; struct page *page;
DEFINE_DMA_UNMAP_ADDR(dma_addr); u32 page_order;
DEFINE_DMA_UNMAP_LEN(len); u32 page_offset;
dma_addr_t dma_addr;
u32 len;
}; };
struct mqnic_ring { struct mqnic_ring {
@ -149,6 +151,7 @@ struct mqnic_ring {
u32 cpl_index; u32 cpl_index;
u32 mtu; u32 mtu;
u32 page_order;
size_t buf_size; size_t buf_size;
u8 *buf; u8 *buf;

View File

@ -66,6 +66,10 @@ static int mqnic_start_port(struct net_device *ndev)
for (k = 0; k < priv->rx_queue_count; k++) for (k = 0; k < priv->rx_queue_count; k++)
{ {
priv->rx_ring[k]->mtu = ndev->mtu; priv->rx_ring[k]->mtu = ndev->mtu;
if (ndev->mtu+ETH_HLEN <= PAGE_SIZE)
priv->rx_ring[k]->page_order = 0;
else
priv->rx_ring[k]->page_order = ilog2((ndev->mtu+ETH_HLEN+PAGE_SIZE-1)/PAGE_SIZE-1)+1;
mqnic_activate_rx_ring(priv, priv->rx_ring[k], k); mqnic_activate_rx_ring(priv, priv->rx_ring[k], k);
} }

View File

@ -166,12 +166,12 @@ void mqnic_rx_write_head_ptr(struct mqnic_ring *ring)
void mqnic_free_rx_desc(struct mqnic_priv *priv, struct mqnic_ring *ring, int index) void mqnic_free_rx_desc(struct mqnic_priv *priv, struct mqnic_ring *ring, int index)
{ {
struct mqnic_rx_info *rx_info = &ring->rx_info[index]; struct mqnic_rx_info *rx_info = &ring->rx_info[index];
struct sk_buff *skb = rx_info->skb; struct page *page = rx_info->page;
dma_unmap_single(priv->dev, dma_unmap_addr(rx_info, dma_addr), dma_unmap_len(rx_info, len), PCI_DMA_FROMDEVICE); dma_unmap_page(priv->dev, dma_unmap_addr(rx_info, dma_addr), dma_unmap_len(rx_info, len), PCI_DMA_FROMDEVICE);
dma_unmap_addr_set(rx_info, dma_addr, 0); rx_info->dma_addr = 0;
napi_consume_skb(skb, 0); __free_pages(page, rx_info->page_order);
rx_info->skb = NULL; rx_info->page = NULL;
} }
int mqnic_free_rx_buf(struct mqnic_priv *priv, struct mqnic_ring *ring) int mqnic_free_rx_buf(struct mqnic_priv *priv, struct mqnic_ring *ring)
@ -198,31 +198,31 @@ int mqnic_prepare_rx_desc(struct mqnic_priv *priv, struct mqnic_ring *ring, int
{ {
struct mqnic_rx_info *rx_info = &ring->rx_info[index]; struct mqnic_rx_info *rx_info = &ring->rx_info[index];
struct mqnic_desc *rx_desc = (struct mqnic_desc *)(ring->buf + index * sizeof(*rx_desc)); struct mqnic_desc *rx_desc = (struct mqnic_desc *)(ring->buf + index * sizeof(*rx_desc));
struct sk_buff *skb = rx_info->skb; struct page *page = rx_info->page;
u32 len = ring->mtu+ETH_HLEN; u32 page_order = ring->page_order;
u32 len = PAGE_SIZE << page_order;
dma_addr_t dma_addr; dma_addr_t dma_addr;
if (unlikely(skb)) if (unlikely(page))
{ {
dev_err(&priv->mdev->pdev->dev, "mqnic_prepare_rx_desc skb not yet processed on port %d", priv->port); dev_err(&priv->mdev->pdev->dev, "mqnic_prepare_rx_desc skb not yet processed on port %d", priv->port);
return -1; return -1;
} }
skb = netdev_alloc_skb_ip_align(priv->ndev, len); page = dev_alloc_pages(page_order);
if (unlikely(!skb)) if (unlikely(!page))
{ {
dev_err(&priv->mdev->pdev->dev, "mqnic_prepare_rx_desc failed to allocate skb on port %d", priv->port); dev_err(&priv->mdev->pdev->dev, "mqnic_prepare_rx_desc failed to allocate memory on port %d", priv->port);
return -1; return -1;
} }
map_skb: // map page
// map skb dma_addr = dma_map_page(priv->dev, page, 0, len, PCI_DMA_FROMDEVICE);
dma_addr = dma_map_single(priv->dev, skb->data, len, PCI_DMA_FROMDEVICE);
if (unlikely(dma_mapping_error(priv->dev, dma_addr))) if (unlikely(dma_mapping_error(priv->dev, dma_addr)))
{ {
dev_err(&priv->mdev->pdev->dev, "mqnic_prepare_rx_desc failed to map skb on port %d", priv->port); dev_err(&priv->mdev->pdev->dev, "mqnic_prepare_rx_desc DMA mapping failed on port %d", priv->port);
napi_consume_skb(rx_info->skb, 0); __free_pages(page, page_order);
return -1; return -1;
} }
@ -231,9 +231,11 @@ map_skb:
rx_desc->addr = dma_addr; rx_desc->addr = dma_addr;
// update rx_info // update rx_info
rx_info->skb = skb; rx_info->page = page;
dma_unmap_addr_set(rx_info, dma_addr, dma_addr); rx_info->page_order = page_order;
dma_unmap_len_set(rx_info, len, len); rx_info->page_offset = 0;
rx_info->dma_addr = dma_addr;
rx_info->len = len;
return 0; return 0;
} }
@ -264,12 +266,14 @@ int mqnic_process_rx_cq(struct net_device *ndev, struct mqnic_cq_ring *cq_ring,
struct mqnic_rx_info *rx_info; struct mqnic_rx_info *rx_info;
struct mqnic_cpl *cpl; struct mqnic_cpl *cpl;
struct sk_buff *skb; struct sk_buff *skb;
struct page *page;
u32 cq_index; u32 cq_index;
u32 cq_tail_ptr; u32 cq_tail_ptr;
u32 ring_index; u32 ring_index;
u32 ring_clean_tail_ptr; u32 ring_clean_tail_ptr;
int done = 0; int done = 0;
int budget = napi_budget; int budget = napi_budget;
u32 len;
if (unlikely(!priv->port_up)) if (unlikely(!priv->port_up))
{ {
@ -290,25 +294,21 @@ int mqnic_process_rx_cq(struct net_device *ndev, struct mqnic_cq_ring *cq_ring,
cpl = (struct mqnic_cpl *)(cq_ring->buf + cq_index * MQNIC_CPL_SIZE); cpl = (struct mqnic_cpl *)(cq_ring->buf + cq_index * MQNIC_CPL_SIZE);
ring_index = cpl->index & ring->size_mask; ring_index = cpl->index & ring->size_mask;
rx_info = &ring->rx_info[ring_index]; rx_info = &ring->rx_info[ring_index];
skb = rx_info->skb; page = rx_info->page;
if (unlikely(!skb)) if (unlikely(!page))
{ {
dev_err(&priv->mdev->pdev->dev, "mqnic_process_rx_cq ring %d null SKB at index %d", cq_ring->ring_index, ring_index); dev_err(&priv->mdev->pdev->dev, "mqnic_process_rx_cq ring %d null page at index %d", cq_ring->ring_index, ring_index);
print_hex_dump(KERN_ERR, "", DUMP_PREFIX_NONE, 16, 1, cpl, MQNIC_CPL_SIZE, true); print_hex_dump(KERN_ERR, "", DUMP_PREFIX_NONE, 16, 1, cpl, MQNIC_CPL_SIZE, true);
break; break;
} }
// set length skb = napi_get_frags(&cq_ring->napi);
if (cpl->len <= skb_tailroom(skb)) if (unlikely(!skb))
{ {
skb_put(skb, cpl->len); dev_err(&priv->mdev->pdev->dev, "mqnic_process_rx_cq ring %d failed to allocate skb", cq_ring->ring_index);
break;
} }
else
{
skb_put(skb, skb_tailroom(skb));
}
skb->protocol = eth_type_trans(skb, priv->ndev);
// RX hardware timestamp // RX hardware timestamp
if (priv->if_features & MQNIC_IF_FEATURE_PTP_TS) if (priv->if_features & MQNIC_IF_FEATURE_PTP_TS)
@ -321,19 +321,31 @@ int mqnic_process_rx_cq(struct net_device *ndev, struct mqnic_cq_ring *cq_ring,
// RX hardware checksum // RX hardware checksum
if ((ndev->features & NETIF_F_RXCSUM) && if ((ndev->features & NETIF_F_RXCSUM) &&
(skb->protocol == htons(ETH_P_IP) || skb->protocol == htons(ETH_P_IPV6)) && (skb->protocol == htons(ETH_P_IP) || skb->protocol == htons(ETH_P_IPV6)) &&
(skb->len >= 64)) { (skb->len >= 64))
{
skb->csum = be16_to_cpu(cpl->rx_csum); skb->csum = be16_to_cpu(cpl->rx_csum);
skb->ip_summed = CHECKSUM_COMPLETE; skb->ip_summed = CHECKSUM_COMPLETE;
} }
// unmap // unmap
dma_unmap_single(priv->dev, dma_unmap_addr(rx_info, dma_addr), dma_unmap_len(rx_info, len), PCI_DMA_FROMDEVICE); dma_unmap_page(priv->dev, dma_unmap_addr(rx_info, dma_addr), dma_unmap_len(rx_info, len), PCI_DMA_FROMDEVICE);
dma_unmap_addr_set(rx_info, dma_addr, 0); rx_info->dma_addr = 0;
len = min_t(u32, cpl->len, rx_info->len);
dma_sync_single_range_for_cpu(priv->dev, rx_info->dma_addr, rx_info->page_offset, rx_info->len, PCI_DMA_FROMDEVICE);
__skb_fill_page_desc(skb, 0, page, rx_info->page_offset, len);
rx_info->page = NULL;
skb_shinfo(skb)->nr_frags = 1;
skb->len = len;
skb->data_len = len;
skb->truesize += rx_info->len;
// hand off SKB // hand off SKB
napi_gro_receive(&cq_ring->napi, skb); napi_gro_frags(&cq_ring->napi);
rx_info->skb = NULL;
ring->packets++; ring->packets++;
ring->bytes += cpl->len; ring->bytes += cpl->len;
@ -359,7 +371,7 @@ int mqnic_process_rx_cq(struct net_device *ndev, struct mqnic_cq_ring *cq_ring,
{ {
rx_info = &ring->rx_info[ring_index]; rx_info = &ring->rx_info[ring_index];
if (rx_info->skb) if (rx_info->page)
break; break;
ring_clean_tail_ptr++; ring_clean_tail_ptr++;