Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next

Resolve conflict between commit 264640fc2c5f4f ("ipv6: distinguish frag
queues by device for multicast and link-local packets") from the net
tree and commit 029f7f3b8701c ("netfilter: ipv6: nf_defrag: avoid/free
clone operations") from the nf-next tree.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Conflicts:
	net/ipv6/netfilter/nf_conntrack_reasm.c

13 files changed, 264 insertions, 114 deletions

diff --git a/drivers/net/ethernet/chelsio/Kconfig b/drivers/net/ethernet/chelsio/Kconfig
index a79813a17b6e..4d187f22c48b 100644
--- a/drivers/net/ethernet/chelsio/Kconfig
+++ b/drivers/net/ethernet/chelsio/Kconfig
@@ -65,13 +65,14 @@ config CHELSIO_T3
 	  will be called cxgb3.
 
 config CHELSIO_T4
-	tristate "Chelsio Communications T4/T5 Ethernet support"
+	tristate "Chelsio Communications T4/T5/T6 Ethernet support"
 	depends on PCI && (IPV6 || IPV6=n)
 	select FW_LOADER
 	select MDIO
 	---help---
-	  This driver supports Chelsio T4 and T5 based gigabit, 10Gb Ethernet
-	  adapter and T5 based 40Gb Ethernet adapter.
+	  This driver supports Chelsio T4, T5 & T6 based gigabit, 10Gb Ethernet
+	  adapter and T5/T6 based 40Gb and T6 based 25Gb, 50Gb and 100Gb
+	  Ethernet adapters.
 
 	  For general information about Chelsio and our products, visit
 	  our website at <http://www.chelsio.com>.
@@ -85,7 +86,7 @@ config CHELSIO_T4
 	  will be called cxgb4.
 
 config CHELSIO_T4_DCB
-	bool "Data Center Bridging (DCB) Support for Chelsio T4/T5 cards"
+	bool "Data Center Bridging (DCB) Support for Chelsio T4/T5/T6 cards"
 	default n
 	depends on CHELSIO_T4 && DCB
 	---help---
@@ -107,12 +108,12 @@ config CHELSIO_T4_FCOE
 	  If unsure, say N.
 
 config CHELSIO_T4VF
-	tristate "Chelsio Communications T4/T5 Virtual Function Ethernet support"
+	tristate "Chelsio Communications T4/T5/T6 Virtual Function Ethernet support"
 	depends on PCI
 	---help---
-	  This driver supports Chelsio T4 and T5 based gigabit, 10Gb Ethernet
-	  adapters and T5 based 40Gb Ethernet adapters with PCI-E SR-IOV Virtual
-	  Functions.
+	  This driver supports Chelsio T4, T5 & T6 based gigabit, 10Gb Ethernet
+	  adapters and T5/T6 based 40Gb and T6 based 25Gb, 50Gb and 100Gb
+	  Ethernet adapters with PCI-E SR-IOV Virtual Functions.
 
 	  For general information about Chelsio and our products, visit
 	  our website at <http://www.chelsio.com>.
diff --git a/drivers/net/ethernet/chelsio/cxgb/pm3393.c b/drivers/net/ethernet/chelsio/cxgb/pm3393.c
index ec5e05052d99..eb462d7db427 100644
--- a/drivers/net/ethernet/chelsio/cxgb/pm3393.c
+++ b/drivers/net/ethernet/chelsio/cxgb/pm3393.c
@@ -570,7 +570,7 @@ static void pm3393_destroy(struct cmac *cmac)
 	kfree(cmac);
 }
 
-static struct cmac_ops pm3393_ops = {
+static const struct cmac_ops pm3393_ops = {
 	.destroy                 = pm3393_destroy,
 	.reset                   = pm3393_reset,
 	.interrupt_enable        = pm3393_interrupt_enable,
diff --git a/drivers/net/ethernet/chelsio/cxgb/vsc7326.c b/drivers/net/ethernet/chelsio/cxgb/vsc7326.c
index b0cb388f5e12..6f30b6f78553 100644
--- a/drivers/net/ethernet/chelsio/cxgb/vsc7326.c
+++ b/drivers/net/ethernet/chelsio/cxgb/vsc7326.c
@@ -666,7 +666,7 @@ static void mac_destroy(struct cmac *mac)
 	kfree(mac);
 }
 
-static struct cmac_ops vsc7326_ops = {
+static const struct cmac_ops vsc7326_ops = {
 	.destroy                  = mac_destroy,
 	.reset                    = mac_reset,
 	.interrupt_handler        = mac_intr_handler,
diff --git a/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c b/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c
index 8f7aa53a4c4b..60908eab3b3a 100644
--- a/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb3/cxgb3_main.c
@@ -701,15 +701,16 @@ static ssize_t attr_store(struct device *d,
 			  ssize_t(*set) (struct net_device *, unsigned int),
 			  unsigned int min_val, unsigned int max_val)
 {
-	char *endp;
 	ssize_t ret;
 	unsigned int val;
 
 	if (!capable(CAP_NET_ADMIN))
 		return -EPERM;
 
-	val = simple_strtoul(buf, &endp, 0);
-	if (endp == buf || val < min_val || val > max_val)
+	ret = kstrtouint(buf, 0, &val);
+	if (ret)
+		return ret;
+	if (val < min_val || val > max_val)
 		return -EINVAL;
 
 	rtnl_lock();
@@ -829,14 +830,15 @@ static ssize_t tm_attr_store(struct device *d,
 	struct port_info *pi = netdev_priv(to_net_dev(d));
 	struct adapter *adap = pi->adapter;
 	unsigned int val;
-	char *endp;
 	ssize_t ret;
 
 	if (!capable(CAP_NET_ADMIN))
 		return -EPERM;
 
-	val = simple_strtoul(buf, &endp, 0);
-	if (endp == buf || val > 10000000)
+	ret = kstrtouint(buf, 0, &val);
+	if (ret)
+		return ret;
+	if (val > 10000000)
 		return -EINVAL;
 
 	rtnl_lock();
diff --git a/drivers/net/ethernet/chelsio/cxgb3/t3_hw.c b/drivers/net/ethernet/chelsio/cxgb3/t3_hw.c
index a22768c94200..ee04caa6c4d8 100644
--- a/drivers/net/ethernet/chelsio/cxgb3/t3_hw.c
+++ b/drivers/net/ethernet/chelsio/cxgb3/t3_hw.c
@@ -709,11 +709,21 @@ static int get_vpd_params(struct adapter *adapter, struct vpd_params *p)
 			return ret;
 	}
 
-	p->cclk = simple_strtoul(vpd.cclk_data, NULL, 10);
-	p->mclk = simple_strtoul(vpd.mclk_data, NULL, 10);
-	p->uclk = simple_strtoul(vpd.uclk_data, NULL, 10);
-	p->mdc = simple_strtoul(vpd.mdc_data, NULL, 10);
-	p->mem_timing = simple_strtoul(vpd.mt_data, NULL, 10);
+	ret = kstrtouint(vpd.cclk_data, 10, &p->cclk);
+	if (ret)
+		return ret;
+	ret = kstrtouint(vpd.mclk_data, 10, &p->mclk);
+	if (ret)
+		return ret;
+	ret = kstrtouint(vpd.uclk_data, 10, &p->uclk);
+	if (ret)
+		return ret;
+	ret = kstrtouint(vpd.mdc_data, 10, &p->mdc);
+	if (ret)
+		return ret;
+	ret = kstrtouint(vpd.mt_data, 10, &p->mem_timing);
+	if (ret)
+		return ret;
 	memcpy(p->sn, vpd.sn_data, SERNUM_LEN);
 
 	/* Old eeproms didn't have port information */
@@ -723,8 +733,12 @@ static int get_vpd_params(struct adapter *adapter, struct vpd_params *p)
 	} else {
 		p->port_type[0] = hex_to_bin(vpd.port0_data[0]);
 		p->port_type[1] = hex_to_bin(vpd.port1_data[0]);
-		p->xauicfg[0] = simple_strtoul(vpd.xaui0cfg_data, NULL, 16);
-		p->xauicfg[1] = simple_strtoul(vpd.xaui1cfg_data, NULL, 16);
+		ret = kstrtou16(vpd.xaui0cfg_data, 16, &p->xauicfg[0]);
+		if (ret)
+			return ret;
+		ret = kstrtou16(vpd.xaui1cfg_data, 16, &p->xauicfg[1]);
+		if (ret)
+			return ret;
 	}
 
 	ret = hex2bin(p->eth_base, vpd.na_data, 6);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/clip_tbl.c b/drivers/net/ethernet/chelsio/cxgb4/clip_tbl.c
index c308429dd9c7..d288dcf6062f 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/clip_tbl.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/clip_tbl.c
@@ -118,6 +118,11 @@ int cxgb4_clip_get(const struct net_device *dev, const u32 *lip, u8 v6)
 			ret = clip6_get_mbox(dev, (const struct in6_addr *)lip);
 			if (ret) {
 				write_unlock_bh(&ctbl->lock);
+				dev_err(adap->pdev_dev,
+					"CLIP FW cmd failed with error %d, "
+					"Connections using %pI6c wont be "
+					"offloaded",
+					ret, ce->addr6.sin6_addr.s6_addr);
 				return ret;
 			}
 		} else {
@@ -127,6 +132,9 @@ int cxgb4_clip_get(const struct net_device *dev, const u32 *lip, u8 v6)
 		}
 	} else {
 		write_unlock_bh(&ctbl->lock);
+		dev_info(adap->pdev_dev, "CLIP table overflow, "
+			 "Connections using %pI6c wont be offloaded",
+			 (void *)lip);
 		return -ENOMEM;
 	}
 	write_unlock_bh(&ctbl->lock);
@@ -146,6 +154,9 @@ void cxgb4_clip_release(const struct net_device *dev, const u32 *lip, u8 v6)
 	int hash;
 	int ret = -1;
 
+	if (!ctbl)
+		return;
+
 	hash = clip_addr_hash(ctbl, addr, v6);
 
 	read_lock_bh(&ctbl->lock);
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index 55a47de544ea..e01e7228f607 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -483,6 +483,8 @@ struct sge_fl {                     /* SGE free-buffer queue state */
 	unsigned int pidx;          /* producer index */
 	unsigned long alloc_failed; /* # of times buffer allocation failed */
 	unsigned long large_alloc_failed;
+	unsigned long mapping_err;  /* # of RX Buffer DMA Mapping failures */
+	unsigned long low;          /* # of times momentarily starving */
 	unsigned long starving;
 	/* RO fields */
 	unsigned int cntxt_id;      /* SGE context id for the free list */
@@ -618,6 +620,7 @@ struct sge_ofld_txq {               /* state for an SGE offload Tx queue */
 	struct adapter *adap;
 	struct sk_buff_head sendq;  /* list of backpressured packets */
 	struct tasklet_struct qresume_tsk; /* restarts the queue */
+	bool service_ofldq_running; /* service_ofldq() is processing sendq */
 	u8 full;                    /* the Tx ring is full */
 	unsigned long mapping_err;  /* # of I/O MMU packet mapping errors */
 } ____cacheline_aligned_in_smp;
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
index 4269944c5db5..0d579b192350 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_debugfs.c
@@ -2325,6 +2325,8 @@ do { \
 		TL("TxMapErr:", mapping_err);
 		RL("FLAllocErr:", fl.alloc_failed);
 		RL("FLLrgAlcErr:", fl.large_alloc_failed);
+		RL("FLMapErr:", fl.mapping_err);
+		RL("FLLow:", fl.low);
 		RL("FLStarving:", fl.starving);
 
 	} else if (iscsi_idx < iscsi_entries) {
@@ -2359,6 +2361,8 @@ do { \
 		RL("RxNoMem:", stats.nomem);
 		RL("FLAllocErr:", fl.alloc_failed);
 		RL("FLLrgAlcErr:", fl.large_alloc_failed);
+		RL("FLMapErr:", fl.mapping_err);
+		RL("FLLow:", fl.low);
 		RL("FLStarving:", fl.starving);
 
 	} else if (rdma_idx < rdma_entries) {
@@ -2388,6 +2392,8 @@ do { \
 		RL("RxNoMem:", stats.nomem);
 		RL("FLAllocErr:", fl.alloc_failed);
 		RL("FLLrgAlcErr:", fl.large_alloc_failed);
+		RL("FLMapErr:", fl.mapping_err);
+		RL("FLLow:", fl.low);
 		RL("FLStarving:", fl.starving);
 
 	} else if (ciq_idx < ciq_entries) {
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_ethtool.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_ethtool.c
index a077f9476daf..2a61a42ab033 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_ethtool.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_ethtool.c
@@ -35,79 +35,79 @@ static void set_msglevel(struct net_device *dev, u32 val)
 }
 
 static const char stats_strings[][ETH_GSTRING_LEN] = {
-	"tx_octets_ok		",
-	"tx_frames_ok		",
-	"tx_broadcast_frames	",
-	"tx_multicast_frames	",
-	"tx_unicast_frames	",
-	"tx_error_frames	",
-
-	"tx_frames_64		",
-	"tx_frames_65_to_127	",
-	"tx_frames_128_to_255	",
-	"tx_frames_256_to_511	",
-	"tx_frames_512_to_1023	",
-	"tx_frames_1024_to_1518	",
-	"tx_frames_1519_to_max	",
-
-	"tx_frames_dropped	",
-	"tx_pause_frames	",
-	"tx_ppp0_frames		",
-	"tx_ppp1_frames		",
-	"tx_ppp2_frames		",
-	"tx_ppp3_frames		",
-	"tx_ppp4_frames		",
-	"tx_ppp5_frames		",
-	"tx_ppp6_frames		",
-	"tx_ppp7_frames		",
-
-	"rx_octets_ok		",
-	"rx_frames_ok		",
-	"rx_broadcast_frames	",
-	"rx_multicast_frames	",
-	"rx_unicast_frames	",
-
-	"rx_frames_too_long	",
-	"rx_jabber_errors	",
-	"rx_fcs_errors		",
-	"rx_length_errors	",
-	"rx_symbol_errors	",
-	"rx_runt_frames		",
-
-	"rx_frames_64		",
-	"rx_frames_65_to_127	",
-	"rx_frames_128_to_255	",
-	"rx_frames_256_to_511	",
-	"rx_frames_512_to_1023	",
-	"rx_frames_1024_to_1518	",
-	"rx_frames_1519_to_max	",
-
-	"rx_pause_frames	",
-	"rx_ppp0_frames		",
-	"rx_ppp1_frames		",
-	"rx_ppp2_frames		",
-	"rx_ppp3_frames		",
-	"rx_ppp4_frames		",
-	"rx_ppp5_frames		",
-	"rx_ppp6_frames		",
-	"rx_ppp7_frames		",
-
-	"rx_bg0_frames_dropped	",
-	"rx_bg1_frames_dropped	",
-	"rx_bg2_frames_dropped	",
-	"rx_bg3_frames_dropped	",
-	"rx_bg0_frames_trunc	",
-	"rx_bg1_frames_trunc	",
-	"rx_bg2_frames_trunc	",
-	"rx_bg3_frames_trunc	",
-
-	"tso			",
-	"tx_csum_offload	",
-	"rx_csum_good		",
-	"vlan_extractions	",
-	"vlan_insertions	",
-	"gro_packets		",
-	"gro_merged		",
+	"tx_octets_ok           ",
+	"tx_frames_ok           ",
+	"tx_broadcast_frames    ",
+	"tx_multicast_frames    ",
+	"tx_unicast_frames      ",
+	"tx_error_frames        ",
+
+	"tx_frames_64           ",
+	"tx_frames_65_to_127    ",
+	"tx_frames_128_to_255   ",
+	"tx_frames_256_to_511   ",
+	"tx_frames_512_to_1023  ",
+	"tx_frames_1024_to_1518 ",
+	"tx_frames_1519_to_max  ",
+
+	"tx_frames_dropped      ",
+	"tx_pause_frames        ",
+	"tx_ppp0_frames         ",
+	"tx_ppp1_frames         ",
+	"tx_ppp2_frames         ",
+	"tx_ppp3_frames         ",
+	"tx_ppp4_frames         ",
+	"tx_ppp5_frames         ",
+	"tx_ppp6_frames         ",
+	"tx_ppp7_frames         ",
+
+	"rx_octets_ok           ",
+	"rx_frames_ok           ",
+	"rx_broadcast_frames    ",
+	"rx_multicast_frames    ",
+	"rx_unicast_frames      ",
+
+	"rx_frames_too_long     ",
+	"rx_jabber_errors       ",
+	"rx_fcs_errors          ",
+	"rx_length_errors       ",
+	"rx_symbol_errors       ",
+	"rx_runt_frames         ",
+
+	"rx_frames_64           ",
+	"rx_frames_65_to_127    ",
+	"rx_frames_128_to_255   ",
+	"rx_frames_256_to_511   ",
+	"rx_frames_512_to_1023  ",
+	"rx_frames_1024_to_1518 ",
+	"rx_frames_1519_to_max  ",
+
+	"rx_pause_frames        ",
+	"rx_ppp0_frames         ",
+	"rx_ppp1_frames         ",
+	"rx_ppp2_frames         ",
+	"rx_ppp3_frames         ",
+	"rx_ppp4_frames         ",
+	"rx_ppp5_frames         ",
+	"rx_ppp6_frames         ",
+	"rx_ppp7_frames         ",
+
+	"rx_bg0_frames_dropped  ",
+	"rx_bg1_frames_dropped  ",
+	"rx_bg2_frames_dropped  ",
+	"rx_bg3_frames_dropped  ",
+	"rx_bg0_frames_trunc    ",
+	"rx_bg1_frames_trunc    ",
+	"rx_bg2_frames_trunc    ",
+	"rx_bg3_frames_trunc    ",
+
+	"tso                    ",
+	"tx_csum_offload        ",
+	"rx_csum_good           ",
+	"vlan_extractions       ",
+	"vlan_insertions        ",
+	"gro_packets            ",
+	"gro_merged             ",
 };
 
 static char adapter_stats_strings[][ETH_GSTRING_LEN] = {
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
index 0d147610a06f..edd706e739fb 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c
@@ -4865,15 +4865,25 @@ static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	}
 
 #if IS_ENABLED(CONFIG_IPV6)
-	adapter->clipt = t4_init_clip_tbl(adapter->clipt_start,
-					  adapter->clipt_end);
-	if (!adapter->clipt) {
-		/* We tolerate a lack of clip_table, giving up
-		 * some functionality
+	if ((CHELSIO_CHIP_VERSION(adapter->params.chip) <= CHELSIO_T5) &&
+	    (!(t4_read_reg(adapter, LE_DB_CONFIG_A) & ASLIPCOMPEN_F))) {
+		/* CLIP functionality is not present in hardware,
+		 * hence disable all offload features
 		 */
 		dev_warn(&pdev->dev,
-			 "could not allocate Clip table, continuing\n");
+			 "CLIP not enabled in hardware, continuing\n");
 		adapter->params.offload = 0;
+	} else {
+		adapter->clipt = t4_init_clip_tbl(adapter->clipt_start,
+						  adapter->clipt_end);
+		if (!adapter->clipt) {
+			/* We tolerate a lack of clip_table, giving up
+			 * some functionality
+			 */
+			dev_warn(&pdev->dev,
+				 "could not allocate Clip table, continuing\n");
+			adapter->params.offload = 0;
+		}
 	}
 #endif
 	if (is_offload(adapter) && tid_init(&adapter->tids) < 0) {
diff --git a/drivers/net/ethernet/chelsio/cxgb4/sge.c b/drivers/net/ethernet/chelsio/cxgb4/sge.c
index 48d8fbb1c220..8d35ce317f67 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/sge.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/sge.c
@@ -406,7 +406,7 @@ static void free_tx_desc(struct adapter *adap, struct sge_txq *q,
  */
 static inline int reclaimable(const struct sge_txq *q)
 {
-	int hw_cidx = ntohs(q->stat->cidx);
+	int hw_cidx = ntohs(ACCESS_ONCE(q->stat->cidx));
 	hw_cidx -= q->cidx;
 	return hw_cidx < 0 ? hw_cidx + q->size : hw_cidx;
 }
@@ -613,6 +613,7 @@ static unsigned int refill_fl(struct adapter *adap, struct sge_fl *q, int n,
 				       PCI_DMA_FROMDEVICE);
 		if (unlikely(dma_mapping_error(adap->pdev_dev, mapping))) {
 			__free_pages(pg, s->fl_pg_order);
+			q->mapping_err++;
 			goto out;   /* do not try small pages for this error */
 		}
 		mapping |= RX_LARGE_PG_BUF;
@@ -642,6 +643,7 @@ alloc_small_pages:
 				       PCI_DMA_FROMDEVICE);
 		if (unlikely(dma_mapping_error(adap->pdev_dev, mapping))) {
 			put_page(pg);
+			q->mapping_err++;
 			goto out;
 		}
 		*d++ = cpu_to_be64(mapping);
@@ -663,6 +665,7 @@ out:	cred = q->avail - cred;
 
 	if (unlikely(fl_starving(adap, q))) {
 		smp_wmb();
+		q->low++;
 		set_bit(q->cntxt_id - adap->sge.egr_start,
 			adap->sge.starving_fl);
 	}
@@ -1029,6 +1032,30 @@ static void inline_tx_skb(const struct sk_buff *skb, const struct sge_txq *q,
 		*p = 0;
 }
 
+static void *inline_tx_skb_header(const struct sk_buff *skb,
+				  const struct sge_txq *q,  void *pos,
+				  int length)
+{
+	u64 *p;
+	int left = (void *)q->stat - pos;
+
+	if (likely(length <= left)) {
+		memcpy(pos, skb->data, length);
+		pos += length;
+	} else {
+		memcpy(pos, skb->data, left);
+		memcpy(q->desc, skb->data + left, length - left);
+		pos = (void *)q->desc + (length - left);
+	}
+	/* 0-pad to multiple of 16 */
+	p = PTR_ALIGN(pos, 8);
+	if ((uintptr_t)p & 8) {
+		*p = 0;
+		return p + 1;
+	}
+	return p;
+}
+
 /*
  * Figure out what HW csum a packet wants and return the appropriate control
  * bits.
@@ -1320,7 +1347,7 @@ out_free:	dev_kfree_skb_any(skb);
  */
 static inline void reclaim_completed_tx_imm(struct sge_txq *q)
 {
-	int hw_cidx = ntohs(q->stat->cidx);
+	int hw_cidx = ntohs(ACCESS_ONCE(q->stat->cidx));
 	int reclaim = hw_cidx - q->cidx;
 
 	if (reclaim < 0)
@@ -1542,24 +1569,50 @@ static void ofldtxq_stop(struct sge_ofld_txq *q, struct sk_buff *skb)
 }
 
 /**
- *	service_ofldq - restart a suspended offload queue
+ *	service_ofldq - service/restart a suspended offload queue
  *	@q: the offload queue
  *
- *	Services an offload Tx queue by moving packets from its packet queue
- *	to the HW Tx ring.  The function starts and ends with the queue locked.
+ *	Services an offload Tx queue by moving packets from its Pending Send
+ *	Queue to the Hardware TX ring.  The function starts and ends with the
+ *	Send Queue locked, but drops the lock while putting the skb at the
+ *	head of the Send Queue onto the Hardware TX Ring.  Dropping the lock
+ *	allows more skbs to be added to the Send Queue by other threads.
+ *	The packet being processed at the head of the Pending Send Queue is
+ *	left on the queue in case we experience DMA Mapping errors, etc.
+ *	and need to give up and restart later.
+ *
+ *	service_ofldq() can be thought of as a task which opportunistically
+ *	uses other threads execution contexts.  We use the Offload Queue
+ *	boolean "service_ofldq_running" to make sure that only one instance
+ *	is ever running at a time ...
  */
 static void service_ofldq(struct sge_ofld_txq *q)
 {
-	u64 *pos;
+	u64 *pos, *before, *end;
 	int credits;
 	struct sk_buff *skb;
+	struct sge_txq *txq;
+	unsigned int left;
 	unsigned int written = 0;
 	unsigned int flits, ndesc;
 
+	/* If another thread is currently in service_ofldq() processing the
+	 * Pending Send Queue then there's nothing to do. Otherwise, flag
+	 * that we're doing the work and continue.  Examining/modifying
+	 * the Offload Queue boolean "service_ofldq_running" must be done
+	 * while holding the Pending Send Queue Lock.
+	 */
+	if (q->service_ofldq_running)
+		return;
+	q->service_ofldq_running = true;
+
 	while ((skb = skb_peek(&q->sendq)) != NULL && !q->full) {
-		/*
-		 * We drop the lock but leave skb on sendq, thus retaining
-		 * exclusive access to the state of the queue.
+		/* We drop the lock while we're working with the skb at the
+		 * head of the Pending Send Queue.  This allows more skbs to
+		 * be added to the Pending Send Queue while we're working on
+		 * this one.  We don't need to lock to guard the TX Ring
+		 * updates because only one thread of execution is ever
+		 * allowed into service_ofldq() at a time.
 		 */
 		spin_unlock(&q->sendq.lock);
 
@@ -1583,9 +1636,32 @@ static void service_ofldq(struct sge_ofld_txq *q)
 		} else {
 			int last_desc, hdr_len = skb_transport_offset(skb);
 
-			memcpy(pos, skb->data, hdr_len);
-			write_sgl(skb, &q->q, (void *)pos + hdr_len,
-				  pos + flits, hdr_len,
+			/* The WR headers  may not fit within one descriptor.
+			 * So we need to deal with wrap-around here.
+			 */
+			before = (u64 *)pos;
+			end = (u64 *)pos + flits;
+			txq = &q->q;
+			pos = (void *)inline_tx_skb_header(skb, &q->q,
+							   (void *)pos,
+							   hdr_len);
+			if (before > (u64 *)pos) {
+				left = (u8 *)end - (u8 *)txq->stat;
+				end = (void *)txq->desc + left;
+			}
+
+			/* If current position is already at the end of the
+			 * ofld queue, reset the current to point to
+			 * start of the queue and update the end ptr as well.
+			 */
+			if (pos == (u64 *)txq->stat) {
+				left = (u8 *)end - (u8 *)txq->stat;
+				end = (void *)txq->desc + left;
+				pos = (void *)txq->desc;
+			}
+
+			write_sgl(skb, &q->q, (void *)pos,
+				  end, hdr_len,
 				  (dma_addr_t *)skb->head);
 #ifdef CONFIG_NEED_DMA_MAP_STATE
 			skb->dev = q->adap->port[0];
@@ -1604,6 +1680,11 @@ static void service_ofldq(struct sge_ofld_txq *q)
 			written = 0;
 		}
 
+		/* Reacquire the Pending Send Queue Lock so we can unlink the
+		 * skb we've just successfully transferred to the TX Ring and
+		 * loop for the next skb which may be at the head of the
+		 * Pending Send Queue.
+		 */
 		spin_lock(&q->sendq.lock);
 		__skb_unlink(skb, &q->sendq);
 		if (is_ofld_imm(skb))
@@ -1611,6 +1692,11 @@ static void service_ofldq(struct sge_ofld_txq *q)
 	}
 	if (likely(written))
 		ring_tx_db(q->adap, &q->q, written);
+
+	/*Indicate that no thread is processing the Pending Send Queue
+	 * currently.
+	 */
+	q->service_ofldq_running = false;
 }
 
 /**
@@ -1624,9 +1710,19 @@ static int ofld_xmit(struct sge_ofld_txq *q, struct sk_buff *skb)
 {
 	skb->priority = calc_tx_flits_ofld(skb);       /* save for restart */
 	spin_lock(&q->sendq.lock);
+
+	/* Queue the new skb onto the Offload Queue's Pending Send Queue.  If
+	 * that results in this new skb being the only one on the queue, start
+	 * servicing it.  If there are other skbs already on the list, then
+	 * either the queue is currently being processed or it's been stopped
+	 * for some reason and it'll be restarted at a later time.  Restart
+	 * paths are triggered by events like experiencing a DMA Mapping Error
+	 * or filling the Hardware TX Ring.
+	 */
 	__skb_queue_tail(&q->sendq, skb);
 	if (q->sendq.qlen == 1)
 		service_ofldq(q);
+
 	spin_unlock(&q->sendq.lock);
 	return NET_XMIT_SUCCESS;
 }
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_pci_id_tbl.h b/drivers/net/ethernet/chelsio/cxgb4/t4_pci_id_tbl.h
index 03ed00c49823..a8dda635456d 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_pci_id_tbl.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_pci_id_tbl.h
@@ -162,6 +162,9 @@ CH_PCI_DEVICE_ID_TABLE_DEFINE_BEGIN
 	CH_PCI_ID_TABLE_FENTRY(0x5095),	/* Custom T540-CR-SO */
 	CH_PCI_ID_TABLE_FENTRY(0x5096),	/* Custom T580-CR */
 	CH_PCI_ID_TABLE_FENTRY(0x5097),	/* Custom T520-KR */
+	CH_PCI_ID_TABLE_FENTRY(0x5098),	/* Custom 2x40G QSFP */
+	CH_PCI_ID_TABLE_FENTRY(0x5099),	/* Custom 2x40G QSFP */
+	CH_PCI_ID_TABLE_FENTRY(0x509a),	/* Custom T520-CR */
 
 	/* T6 adapters:
 	 */
diff --git a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
index fc3044c8ac1c..91b52a21a2e7 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/t4_regs.h
@@ -2802,6 +2802,10 @@
 #define HASHEN_V(x) ((x) << HASHEN_S)
 #define HASHEN_F    HASHEN_V(1U)
 
+#define ASLIPCOMPEN_S    17
+#define ASLIPCOMPEN_V(x) ((x) << ASLIPCOMPEN_S)
+#define ASLIPCOMPEN_F    ASLIPCOMPEN_V(1U)
+
 #define REQQPARERR_S    16
 #define REQQPARERR_V(x) ((x) << REQQPARERR_S)
 #define REQQPARERR_F    REQQPARERR_V(1U)