From: Ganesh Goudar <ganeshgr@chelsio.com>
Date: Fri, 11 May 2018 18:35:33 +0530
Subject: cxgb4: enable inner header checksum calculation
Patch-mainline: v4.18-rc1
Git-commit: c50ae55e41ee226b96cf204cd1409bc0057b484e
References: bsc#1097585 bsc#1097586 bsc#1097587 bsc#1097588 bsc#1097583 bsc#1097584
set cntrl bits to indicate whether inner header checksum
needs to be calculated whenever the packet is an encapsulated
packet and enable supported encap features.
Fixes: d0a1299c6bf7 ("cxgb4: add support for vxlan segmentation offload")
Signed-off-by: Ganesh Goudar <ganeshgr@chelsio.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 9 ++
drivers/net/ethernet/chelsio/cxgb4/sge.c | 86 ++++++++++++++++++------
drivers/net/ethernet/chelsio/cxgb4/t4_msg.h | 5 +
3 files changed, 81 insertions(+), 19 deletions(-)
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -5617,8 +5617,15 @@ static int init_one(struct pci_dev *pdev
NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
NETIF_F_HW_TC;
- if (CHELSIO_CHIP_VERSION(chip) > CHELSIO_T5)
+ if (CHELSIO_CHIP_VERSION(chip) > CHELSIO_T5) {
+ netdev->hw_enc_features |= NETIF_F_IP_CSUM |
+ NETIF_F_IPV6_CSUM |
+ NETIF_F_RXCSUM |
+ NETIF_F_GSO_UDP_TUNNEL |
+ NETIF_F_TSO | NETIF_F_TSO6;
+
netdev->hw_features |= NETIF_F_GSO_UDP_TUNNEL;
+ }
if (highdma)
netdev->hw_features |= NETIF_F_HIGHDMA;
--- a/drivers/net/ethernet/chelsio/cxgb4/sge.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/sge.c
@@ -1072,12 +1072,27 @@ static void *inline_tx_skb_header(const
static u64 hwcsum(enum chip_type chip, const struct sk_buff *skb)
{
int csum_type;
- const struct iphdr *iph = ip_hdr(skb);
+ bool inner_hdr_csum = false;
+ u16 proto, ver;
- if (iph->version == 4) {
- if (iph->protocol == IPPROTO_TCP)
+ if (skb->encapsulation &&
+ (CHELSIO_CHIP_VERSION(chip) > CHELSIO_T5))
+ inner_hdr_csum = true;
+
+ if (inner_hdr_csum) {
+ ver = inner_ip_hdr(skb)->version;
+ proto = (ver == 4) ? inner_ip_hdr(skb)->protocol :
+ inner_ipv6_hdr(skb)->nexthdr;
+ } else {
+ ver = ip_hdr(skb)->version;
+ proto = (ver == 4) ? ip_hdr(skb)->protocol :
+ ipv6_hdr(skb)->nexthdr;
+ }
+
+ if (ver == 4) {
+ if (proto == IPPROTO_TCP)
csum_type = TX_CSUM_TCPIP;
- else if (iph->protocol == IPPROTO_UDP)
+ else if (proto == IPPROTO_UDP)
csum_type = TX_CSUM_UDPIP;
else {
nocsum: /*
@@ -1090,19 +1105,29 @@ nocsum: /*
/*
* this doesn't work with extension headers
*/
- const struct ipv6hdr *ip6h = (const struct ipv6hdr *)iph;
-
- if (ip6h->nexthdr == IPPROTO_TCP)
+ if (proto == IPPROTO_TCP)
csum_type = TX_CSUM_TCPIP6;
- else if (ip6h->nexthdr == IPPROTO_UDP)
+ else if (proto == IPPROTO_UDP)
csum_type = TX_CSUM_UDPIP6;
else
goto nocsum;
}
if (likely(csum_type >= TX_CSUM_TCPIP)) {
- u64 hdr_len = TXPKT_IPHDR_LEN_V(skb_network_header_len(skb));
- int eth_hdr_len = skb_network_offset(skb) - ETH_HLEN;
+ int eth_hdr_len, l4_len;
+ u64 hdr_len;
+
+ if (inner_hdr_csum) {
+ /* This allows checksum offload for all encapsulated
+ * packets like GRE etc..
+ */
+ l4_len = skb_inner_network_header_len(skb);
+ eth_hdr_len = skb_inner_network_offset(skb) - ETH_HLEN;
+ } else {
+ l4_len = skb_network_header_len(skb);
+ eth_hdr_len = skb_network_offset(skb) - ETH_HLEN;
+ }
+ hdr_len = TXPKT_IPHDR_LEN_V(l4_len);
if (CHELSIO_CHIP_VERSION(chip) <= CHELSIO_T5)
hdr_len |= TXPKT_ETHHDR_LEN_V(eth_hdr_len);
@@ -1273,7 +1298,7 @@ static inline void t6_fill_tnl_lso(struc
netdev_tx_t t4_eth_xmit(struct sk_buff *skb, struct net_device *dev)
{
u32 wr_mid, ctrl0, op;
- u64 cntrl, *end;
+ u64 cntrl, *end, *sgl;
int qidx, credits;
unsigned int flits, ndesc;
struct adapter *adap;
@@ -1443,6 +1468,19 @@ out_free: dev_kfree_skb_any(skb);
TX_CSUM_TCPIP6 : TX_CSUM_TCPIP) |
TXPKT_IPHDR_LEN_V(l3hdr_len);
}
+ sgl = (u64 *)(cpl + 1); /* sgl start here */
+ if (unlikely((u8 *)sgl >= (u8 *)q->q.stat)) {
+ /* If current position is already at the end of the
+ * txq, reset the current to point to start of the queue
+ * and update the end ptr as well.
+ */
+ if (sgl == (u64 *)q->q.stat) {
+ int left = (u8 *)end - (u8 *)q->q.stat;
+
+ end = (void *)q->q.desc + left;
+ sgl = (void *)q->q.desc;
+ }
+ }
q->tso++;
q->tx_cso += ssi->gso_segs;
} else {
@@ -1454,6 +1492,7 @@ out_free: dev_kfree_skb_any(skb);
wr->op_immdlen = htonl(FW_WR_OP_V(op) |
FW_WR_IMMDLEN_V(len));
cpl = (void *)(wr + 1);
+ sgl = (u64 *)(cpl + 1);
if (skb->ip_summed == CHECKSUM_PARTIAL) {
cntrl = hwcsum(adap->params.chip, skb) |
TXPKT_IPCSUM_DIS_F;
@@ -2164,7 +2203,7 @@ static void cxgb4_sgetim_to_hwtstamp(str
}
static void do_gro(struct sge_eth_rxq *rxq, const struct pkt_gl *gl,
- const struct cpl_rx_pkt *pkt)
+ const struct cpl_rx_pkt *pkt, unsigned long tnl_hdr_len)
{
struct adapter *adapter = rxq->rspq.adap;
struct sge *s = &adapter->sge;
@@ -2180,6 +2219,8 @@ static void do_gro(struct sge_eth_rxq *r
}
copy_frags(skb, gl, s->pktshift);
+ if (tnl_hdr_len)
+ skb->csum_level = 1;
skb->len = gl->tot_len - s->pktshift;
skb->data_len = skb->len;
skb->truesize += skb->data_len;
@@ -2311,7 +2352,7 @@ int t4_ethrx_handler(struct sge_rspq *q,
struct sge *s = &q->adap->sge;
int cpl_trace_pkt = is_t4(q->adap->params.chip) ?
CPL_TRACE_PKT : CPL_TRACE_PKT_T5;
- u16 err_vec;
+ u16 err_vec, tnl_hdr_len = 0;
struct port_info *pi;
int ret = 0;
@@ -2320,16 +2361,19 @@ int t4_ethrx_handler(struct sge_rspq *q,
pkt = (const struct cpl_rx_pkt *)rsp;
/* Compressed error vector is enabled for T6 only */
- if (q->adap->params.tp.rx_pkt_encap)
+ if (q->adap->params.tp.rx_pkt_encap) {
err_vec = T6_COMPR_RXERR_VEC_G(be16_to_cpu(pkt->err_vec));
- else
+ tnl_hdr_len = T6_RX_TNLHDR_LEN_G(ntohs(pkt->err_vec));
+ } else {
err_vec = be16_to_cpu(pkt->err_vec);
+ }
csum_ok = pkt->csum_calc && !err_vec &&
(q->netdev->features & NETIF_F_RXCSUM);
- if ((pkt->l2info & htonl(RXF_TCP_F)) &&
+ if (((pkt->l2info & htonl(RXF_TCP_F)) ||
+ tnl_hdr_len) &&
(q->netdev->features & NETIF_F_GRO) && csum_ok && !pkt->ip_frag) {
- do_gro(rxq, si, pkt);
+ do_gro(rxq, si, pkt, tnl_hdr_len);
return 0;
}
@@ -2376,7 +2420,13 @@ int t4_ethrx_handler(struct sge_rspq *q,
} else if (pkt->l2info & htonl(RXF_IP_F)) {
__sum16 c = (__force __sum16)pkt->csum;
skb->csum = csum_unfold(c);
- skb->ip_summed = CHECKSUM_COMPLETE;
+
+ if (tnl_hdr_len) {
+ skb->ip_summed = CHECKSUM_UNNECESSARY;
+ skb->csum_level = 1;
+ } else {
+ skb->ip_summed = CHECKSUM_COMPLETE;
+ }
rxq->stats.rx_cso++;
}
} else {
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_msg.h
@@ -1229,6 +1229,11 @@ struct cpl_rx_pkt {
#define T6_COMPR_RXERR_SUM_V(x) ((x) << T6_COMPR_RXERR_SUM_S)
#define T6_COMPR_RXERR_SUM_F T6_COMPR_RXERR_SUM_V(1U)
+#define T6_RX_TNLHDR_LEN_S 8
+#define T6_RX_TNLHDR_LEN_M 0xFF
+#define T6_RX_TNLHDR_LEN_V(x) ((x) << T6_RX_TNLHDR_LEN_S)
+#define T6_RX_TNLHDR_LEN_G(x) (((x) >> T6_RX_TNLHDR_LEN_S) & T6_RX_TNLHDR_LEN_M)
+
struct cpl_trace_pkt {
u8 opcode;
u8 intf;