[net-next 00/13][pull request] Intel Wired LAN Driver Updates

[net-next 00/13][pull request] Intel Wired LAN Driver Updates

[Jeff Kirsher]

This series contains updates to ixgbe only.

The following are changes since commit 72ec301a27badb11635b08dc845101815abcf4a7:
  Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net-next
and are available in the git repository at:
  git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/net-next master

Alexander Duyck (7):
  ixgbe: Add support for IPv6 and UDP to ixgbe_get_headlen
  ixgbe: Add support for tracking the default user priority to SR-IOV
  ixgbe: Add support for GET_QUEUES message to get DCB configuration
  ixgbe: Enable support for VF API version 1.1 in the PF.
  ixgbevf: Add VF DCB + SR-IOV support
  ixgbe: Drop unnecessary addition from ixgbe_set_rx_buffer_len
  ixgbe: Fix possible memory leak in ixgbe_set_ringparam

Don Skidmore (2):
  ixgbe: Add function ixgbe_reset_pipeline_82599
  ixgbe: Add support for pipeline reset

Emil Tantilov (1):
  ixgbe: add WOL support for new subdevice id

Jacob Keller (1):
  ixgbe: (PTP) refactor init, cyclecounter and reset

Tushar Dave (1):
  ixgbe: Correcting small packet padding

Wei Yongjun (1):
  ixgbe: using is_zero_ether_addr() to simplify the code

 drivers/net/ethernet/intel/ixgbe/ixgbe.h          |   7 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c    | 149 ++++++++++++++++----
 drivers/net/ethernet/intel/ixgbe/ixgbe_common.c   |  73 +++++++++-
 drivers/net/ethernet/intel/ixgbe/ixgbe_common.h   |   1 +
 drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c  | 102 +++++++-------
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c     |  66 +++++++--
 drivers/net/ethernet/intel/ixgbe/ixgbe_mbx.h      |  10 ++
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c      | 109 +++++++--------
 drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c    | 124 +++++++++++++----
 drivers/net/ethernet/intel/ixgbe/ixgbe_type.h     |   1 +
 drivers/net/ethernet/intel/ixgbevf/defines.h      |   7 +-
 drivers/net/ethernet/intel/ixgbevf/ixgbevf.h      |   4 +-
 drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c | 159 +++++++++++++++++++++-
 drivers/net/ethernet/intel/ixgbevf/mbx.h          |  10 ++
 drivers/net/ethernet/intel/ixgbevf/vf.c           |  58 ++++++++
 drivers/net/ethernet/intel/ixgbevf/vf.h           |   2 +
 16 files changed, 688 insertions(+), 194 deletions(-)

-- 
1.7.11.7

[net-next 01/13] ixgbe: Add support for IPv6 and UDP to ixgbe_get_headlen

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

This change adds support for IPv6 and UDP to ixgbe_get_headlen. The
advantage to this is that we can now handle ipv4/UDP, ipv6/TCP, and
ipv6/UDP with a single memcpy instead of having to do them in multiple
pskb_may_pull calls.

A quick bit of testing shows that we increase throughput for a single
session of netperf from 8800Mpbs to about 9300Mpbs in the case of ipv6/TCP.
As such overall ipv6 performance should improve with this change.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Stephen Ko  <stephen.s.ko@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 15 ++++++++++++++-
 1 file changed, 14 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index e2a6691..3ef74f8 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -1244,6 +1244,7 @@ static unsigned int ixgbe_get_headlen(unsigned char *data,
 		struct vlan_hdr *vlan;
 		/* l3 headers */
 		struct iphdr *ipv4;
+		struct ipv6hdr *ipv6;
 	} hdr;
 	__be16 protocol;
 	u8 nexthdr = 0;	/* default to not TCP */
@@ -1284,6 +1285,13 @@ static unsigned int ixgbe_get_headlen(unsigned char *data,
 		/* record next protocol */
 		nexthdr = hdr.ipv4->protocol;
 		hdr.network += hlen;
+	} else if (protocol == __constant_htons(ETH_P_IPV6)) {
+		if ((hdr.network - data) > (max_len - sizeof(struct ipv6hdr)))
+			return max_len;
+
+		/* record next protocol */
+		nexthdr = hdr.ipv6->nexthdr;
+		hdr.network += sizeof(struct ipv6hdr);
 #ifdef IXGBE_FCOE
 	} else if (protocol == __constant_htons(ETH_P_FCOE)) {
 		if ((hdr.network - data) > (max_len - FCOE_HEADER_LEN))
@@ -1294,7 +1302,7 @@ static unsigned int ixgbe_get_headlen(unsigned char *data,
 		return hdr.network - data;
 	}
 
-	/* finally sort out TCP */
+	/* finally sort out TCP/UDP */
 	if (nexthdr == IPPROTO_TCP) {
 		if ((hdr.network - data) > (max_len - sizeof(struct tcphdr)))
 			return max_len;
@@ -1307,6 +1315,11 @@ static unsigned int ixgbe_get_headlen(unsigned char *data,
 			return hdr.network - data;
 
 		hdr.network += hlen;
+	} else if (nexthdr == IPPROTO_UDP) {
+		if ((hdr.network - data) > (max_len - sizeof(struct udphdr)))
+			return max_len;
+
+		hdr.network += sizeof(struct udphdr);
 	}
 
 	/*
-- 
1.7.11.7

[net-next 02/13] ixgbe: Add support for tracking the default user priority to SR-IOV

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

It is necessary to track the default user priority in the PF so that we can
force it upon the VFs.  The motivation behind this is to keep the VFs from
getting access to user priorities meant for things like storage.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h       |  2 +
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c  | 20 ++++++++++
 drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c | 53 ++++++++++++++++----------
 3 files changed, 55 insertions(+), 20 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
index ccb8505..101e525 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
@@ -601,6 +601,8 @@ struct ixgbe_adapter {
 #ifdef CONFIG_DEBUG_FS
 	struct dentry *ixgbe_dbg_adapter;
 #endif /*CONFIG_DEBUG_FS*/
+
+	u8 default_up;
 };
 
 struct ixgbe_fdir_filter {
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index 3ef74f8..35be7d3 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -5460,6 +5460,23 @@ static void ixgbe_watchdog_update_link(struct ixgbe_adapter *adapter)
 	adapter->link_speed = link_speed;
 }
 
+static void ixgbe_update_default_up(struct ixgbe_adapter *adapter)
+{
+#ifdef CONFIG_IXGBE_DCB
+	struct net_device *netdev = adapter->netdev;
+	struct dcb_app app = {
+			      .selector = IEEE_8021QAZ_APP_SEL_ETHERTYPE,
+			      .protocol = 0,
+			     };
+	u8 up = 0;
+
+	if (adapter->dcbx_cap & DCB_CAP_DCBX_VER_IEEE)
+		up = dcb_ieee_getapp_mask(netdev, &app);
+
+	adapter->default_up = (up > 1) ? (ffs(up) - 1) : 0;
+#endif
+}
+
 /**
  * ixgbe_watchdog_link_is_up - update netif_carrier status and
  *                             print link up message
@@ -5519,6 +5536,9 @@ static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
 	netif_carrier_on(netdev);
 	ixgbe_check_vf_rate_limit(adapter);
 
+	/* update the default user priority for VFs */
+	ixgbe_update_default_up(adapter);
+
 	/* ping all the active vfs to let them know link has changed */
 	ixgbe_ping_all_vfs(adapter);
 }
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c
index f563625..b330a1c 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c
@@ -431,35 +431,47 @@ static void ixgbe_set_vmolr(struct ixgbe_hw *hw, u32 vf, bool aupe)
 	IXGBE_WRITE_REG(hw, IXGBE_VMOLR(vf), vmolr);
 }
 
-static void ixgbe_set_vmvir(struct ixgbe_adapter *adapter, u32 vid, u32 vf)
+static void ixgbe_set_vmvir(struct ixgbe_adapter *adapter,
+			    u16 vid, u16 qos, u32 vf)
 {
 	struct ixgbe_hw *hw = &adapter->hw;
+	u32 vmvir = vid | (qos << VLAN_PRIO_SHIFT) | IXGBE_VMVIR_VLANA_DEFAULT;
 
-	if (vid)
-		IXGBE_WRITE_REG(hw, IXGBE_VMVIR(vf),
-				(vid | IXGBE_VMVIR_VLANA_DEFAULT));
-	else
-		IXGBE_WRITE_REG(hw, IXGBE_VMVIR(vf), 0);
+	IXGBE_WRITE_REG(hw, IXGBE_VMVIR(vf), vmvir);
 }
 
+static void ixgbe_clear_vmvir(struct ixgbe_adapter *adapter, u32 vf)
+{
+	struct ixgbe_hw *hw = &adapter->hw;
+
+	IXGBE_WRITE_REG(hw, IXGBE_VMVIR(vf), 0);
+}
 static inline void ixgbe_vf_reset_event(struct ixgbe_adapter *adapter, u32 vf)
 {
 	struct ixgbe_hw *hw = &adapter->hw;
+	struct vf_data_storage *vfinfo = &adapter->vfinfo[vf];
 	int rar_entry = hw->mac.num_rar_entries - (vf + 1);
+	u8 num_tcs = netdev_get_num_tc(adapter->netdev);
+
+	/* add PF assigned VLAN or VLAN 0 */
+	ixgbe_set_vf_vlan(adapter, true, vfinfo->pf_vlan, vf);
 
 	/* reset offloads to defaults */
-	if (adapter->vfinfo[vf].pf_vlan) {
-		ixgbe_set_vf_vlan(adapter, true,
-				  adapter->vfinfo[vf].pf_vlan, vf);
-		ixgbe_set_vmvir(adapter,
-				(adapter->vfinfo[vf].pf_vlan |
-				 (adapter->vfinfo[vf].pf_qos <<
-				  VLAN_PRIO_SHIFT)), vf);
-		ixgbe_set_vmolr(hw, vf, false);
+	ixgbe_set_vmolr(hw, vf, !vfinfo->pf_vlan);
+
+	/* set outgoing tags for VFs */
+	if (!vfinfo->pf_vlan && !vfinfo->pf_qos && !num_tcs) {
+		ixgbe_clear_vmvir(adapter, vf);
 	} else {
-		ixgbe_set_vf_vlan(adapter, true, 0, vf);
-		ixgbe_set_vmvir(adapter, 0, vf);
-		ixgbe_set_vmolr(hw, vf, true);
+		if (vfinfo->pf_qos || !num_tcs)
+			ixgbe_set_vmvir(adapter, vfinfo->pf_vlan,
+					vfinfo->pf_qos, vf);
+		else
+			ixgbe_set_vmvir(adapter, vfinfo->pf_vlan,
+					adapter->default_up, vf);
+
+		if (vfinfo->spoofchk_enabled)
+			hw->mac.ops.set_vlan_anti_spoofing(hw, true, vf);
 	}
 
 	/* reset multicast table array for vf */
@@ -661,8 +673,9 @@ static int ixgbe_set_vf_vlan_msg(struct ixgbe_adapter *adapter,
 	int add = (msgbuf[0] & IXGBE_VT_MSGINFO_MASK) >> IXGBE_VT_MSGINFO_SHIFT;
 	int vid = (msgbuf[1] & IXGBE_VLVF_VLANID_MASK);
 	int err;
+	u8 tcs = netdev_get_num_tc(adapter->netdev);
 
-	if (adapter->vfinfo[vf].pf_vlan) {
+	if (adapter->vfinfo[vf].pf_vlan || tcs) {
 		e_warn(drv,
 		       "VF %d attempted to override administratively set VLAN configuration\n"
 		       "Reload the VF driver to resume operations\n",
@@ -896,7 +909,7 @@ int ixgbe_ndo_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
 		err = ixgbe_set_vf_vlan(adapter, true, vlan, vf);
 		if (err)
 			goto out;
-		ixgbe_set_vmvir(adapter, vlan | (qos << VLAN_PRIO_SHIFT), vf);
+		ixgbe_set_vmvir(adapter, vlan, qos, vf);
 		ixgbe_set_vmolr(hw, vf, false);
 		if (adapter->vfinfo[vf].spoofchk_enabled)
 			hw->mac.ops.set_vlan_anti_spoofing(hw, true, vf);
@@ -916,7 +929,7 @@ int ixgbe_ndo_set_vf_vlan(struct net_device *netdev, int vf, u16 vlan, u8 qos)
 	} else {
 		err = ixgbe_set_vf_vlan(adapter, false,
 					adapter->vfinfo[vf].pf_vlan, vf);
-		ixgbe_set_vmvir(adapter, vlan, vf);
+		ixgbe_clear_vmvir(adapter, vf);
 		ixgbe_set_vmolr(hw, vf, true);
 		hw->mac.ops.set_vlan_anti_spoofing(hw, false, vf);
 		if (adapter->vfinfo[vf].vlan_count)
-- 
1.7.11.7

[net-next 03/13] ixgbe: Add support for GET_QUEUES message to get DCB configuration

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

This patch addresses several issues in regards to the combination of DCB
and SR-IOV. Specifically it allows us to send information to the VF on
which queues it should be using.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Tested-by: Ross Brattain <ross.b.brattain@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_mbx.h   | 10 ++++++
 drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c | 42 ++++++++++++++++++++++++++
 2 files changed, 52 insertions(+)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_mbx.h b/drivers/net/ethernet/intel/ixgbe/ixgbe_mbx.h
index d4c842e..42dd65e 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_mbx.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_mbx.h
@@ -71,6 +71,7 @@
 enum ixgbe_pfvf_api_rev {
 	ixgbe_mbox_api_10,	/* API version 1.0, linux/freebsd VF driver */
 	ixgbe_mbox_api_20,	/* API version 2.0, solaris Phase1 VF driver */
+	ixgbe_mbox_api_11,	/* API version 1.1, linux/freebsd VF driver */
 	/* This value should always be last */
 	ixgbe_mbox_api_unknown,	/* indicates that API version is not known */
 };
@@ -86,6 +87,15 @@ enum ixgbe_pfvf_api_rev {
 #define IXGBE_VF_SET_MACVLAN	0x06 /* VF requests PF for unicast filter */
 #define IXGBE_VF_API_NEGOTIATE	0x08 /* negotiate API version */
 
+/* mailbox API, version 1.1 VF requests */
+#define IXGBE_VF_GET_QUEUES	0x09 /* get queue configuration */
+
+/* GET_QUEUES return data indices within the mailbox */
+#define IXGBE_VF_TX_QUEUES	1	/* number of Tx queues supported */
+#define IXGBE_VF_RX_QUEUES	2	/* number of Rx queues supported */
+#define IXGBE_VF_TRANS_VLAN	3	/* Indication of port vlan */
+#define IXGBE_VF_DEF_QUEUE	4	/* Default queue offset */
+
 /* length of permanent address message returned from PF */
 #define IXGBE_VF_PERMADDR_MSG_LEN 4
 /* word in permanent address message with the current multicast type */
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c
index b330a1c..8bf467b 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c
@@ -751,6 +751,45 @@ static int ixgbe_negotiate_vf_api(struct ixgbe_adapter *adapter,
 	return -1;
 }
 
+static int ixgbe_get_vf_queues(struct ixgbe_adapter *adapter,
+			       u32 *msgbuf, u32 vf)
+{
+	struct net_device *dev = adapter->netdev;
+	struct ixgbe_ring_feature *vmdq = &adapter->ring_feature[RING_F_VMDQ];
+	unsigned int default_tc = 0;
+	u8 num_tcs = netdev_get_num_tc(dev);
+
+	/* verify the PF is supporting the correct APIs */
+	switch (adapter->vfinfo[vf].vf_api) {
+	case ixgbe_mbox_api_20:
+	case ixgbe_mbox_api_11:
+		break;
+	default:
+		return -1;
+	}
+
+	/* only allow 1 Tx queue for bandwidth limiting */
+	msgbuf[IXGBE_VF_TX_QUEUES] = __ALIGN_MASK(1, ~vmdq->mask);
+	msgbuf[IXGBE_VF_RX_QUEUES] = __ALIGN_MASK(1, ~vmdq->mask);
+
+	/* if TCs > 1 determine which TC belongs to default user priority */
+	if (num_tcs > 1)
+		default_tc = netdev_get_prio_tc_map(dev, adapter->default_up);
+
+	/* notify VF of need for VLAN tag stripping, and correct queue */
+	if (num_tcs)
+		msgbuf[IXGBE_VF_TRANS_VLAN] = num_tcs;
+	else if (adapter->vfinfo[vf].pf_vlan || adapter->vfinfo[vf].pf_qos)
+		msgbuf[IXGBE_VF_TRANS_VLAN] = 1;
+	else
+		msgbuf[IXGBE_VF_TRANS_VLAN] = 0;
+
+	/* notify VF of default queue */
+	msgbuf[IXGBE_VF_DEF_QUEUE] = default_tc;
+
+	return 0;
+}
+
 static int ixgbe_rcv_msg_from_vf(struct ixgbe_adapter *adapter, u32 vf)
 {
 	u32 mbx_size = IXGBE_VFMAILBOX_SIZE;
@@ -804,6 +843,9 @@ static int ixgbe_rcv_msg_from_vf(struct ixgbe_adapter *adapter, u32 vf)
 	case IXGBE_VF_API_NEGOTIATE:
 		retval = ixgbe_negotiate_vf_api(adapter, msgbuf, vf);
 		break;
+	case IXGBE_VF_GET_QUEUES:
+		retval = ixgbe_get_vf_queues(adapter, msgbuf, vf);
+		break;
 	default:
 		e_err(drv, "Unhandled Msg %8.8x\n", msgbuf[0]);
 		retval = IXGBE_ERR_MBX;
-- 
1.7.11.7

[net-next 04/13] ixgbe: Enable support for VF API version 1.1 in the PF.

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

This change switches on the last few bits for us enabling version 1.1 VF
support in the PF.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Tested-by: Robert Garrett <RobertX.Garrett@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c | 29 +++++++++++++++++++-------
 1 file changed, 21 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c
index 8bf467b..96876b7 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_sriov.c
@@ -371,14 +371,26 @@ static s32 ixgbe_set_vf_lpe(struct ixgbe_adapter *adapter, u32 *msgbuf, u32 vf)
 					     IXGBE_FCOE_JUMBO_FRAME_SIZE);
 
 #endif /* CONFIG_FCOE */
-		/*
-		 * If the PF or VF are running w/ jumbo frames enabled we
-		 * need to shut down the VF Rx path as we cannot support
-		 * jumbo frames on legacy VFs
-		 */
-		if ((pf_max_frame > ETH_FRAME_LEN) ||
-		    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
-			err = -EINVAL;
+		switch (adapter->vfinfo[vf].vf_api) {
+		case ixgbe_mbox_api_11:
+			/*
+			 * Version 1.1 supports jumbo frames on VFs if PF has
+			 * jumbo frames enabled which means legacy VFs are
+			 * disabled
+			 */
+			if (pf_max_frame > ETH_FRAME_LEN)
+				break;
+		default:
+			/*
+			 * If the PF or VF are running w/ jumbo frames enabled
+			 * we need to shut down the VF Rx path as we cannot
+			 * support jumbo frames on legacy VFs
+			 */
+			if ((pf_max_frame > ETH_FRAME_LEN) ||
+			    (max_frame > (ETH_FRAME_LEN + ETH_FCS_LEN)))
+				err = -EINVAL;
+			break;
+		}
 
 		/* determine VF receive enable location */
 		vf_shift = vf % 32;
@@ -740,6 +752,7 @@ static int ixgbe_negotiate_vf_api(struct ixgbe_adapter *adapter,
 
 	switch (api) {
 	case ixgbe_mbox_api_10:
+	case ixgbe_mbox_api_11:
 		adapter->vfinfo[vf].vf_api = api;
 		return 0;
 	default:
-- 
1.7.11.7

[net-next 05/13] ixgbevf: Add VF DCB + SR-IOV support

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

This change adds support for DCB and SR-IOV from the VF.  With this change
in place the VF will correctly use a traffic class other than 0 in the case
that the PF is configured with the default user priority belonging to a
traffic class other than 0.

Cc: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Sibai Li <sibai.li@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbevf/defines.h      |   7 +-
 drivers/net/ethernet/intel/ixgbevf/ixgbevf.h      |   4 +-
 drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c | 159 +++++++++++++++++++++-
 drivers/net/ethernet/intel/ixgbevf/mbx.h          |  10 ++
 drivers/net/ethernet/intel/ixgbevf/vf.c           |  58 ++++++++
 drivers/net/ethernet/intel/ixgbevf/vf.h           |   2 +
 6 files changed, 232 insertions(+), 8 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbevf/defines.h b/drivers/net/ethernet/intel/ixgbevf/defines.h
index da17ccf..3147795 100644
--- a/drivers/net/ethernet/intel/ixgbevf/defines.h
+++ b/drivers/net/ethernet/intel/ixgbevf/defines.h
@@ -33,8 +33,11 @@
 #define IXGBE_DEV_ID_X540_VF            0x1515
 
 #define IXGBE_VF_IRQ_CLEAR_MASK         7
-#define IXGBE_VF_MAX_TX_QUEUES          1
-#define IXGBE_VF_MAX_RX_QUEUES          1
+#define IXGBE_VF_MAX_TX_QUEUES          8
+#define IXGBE_VF_MAX_RX_QUEUES          8
+
+/* DCB define */
+#define IXGBE_VF_MAX_TRAFFIC_CLASS	8
 
 /* Link speed */
 typedef u32 ixgbe_link_speed;
diff --git a/drivers/net/ethernet/intel/ixgbevf/ixgbevf.h b/drivers/net/ethernet/intel/ixgbevf/ixgbevf.h
index 4a9c9c2..2323ccd 100644
--- a/drivers/net/ethernet/intel/ixgbevf/ixgbevf.h
+++ b/drivers/net/ethernet/intel/ixgbevf/ixgbevf.h
@@ -89,8 +89,8 @@ struct ixgbevf_ring {
 /* How many Rx Buffers do we bundle into one write to the hardware ? */
 #define IXGBEVF_RX_BUFFER_WRITE	16	/* Must be power of 2 */
 
-#define MAX_RX_QUEUES 1
-#define MAX_TX_QUEUES 1
+#define MAX_RX_QUEUES IXGBE_VF_MAX_RX_QUEUES
+#define MAX_TX_QUEUES IXGBE_VF_MAX_TX_QUEUES
 
 #define IXGBEVF_DEFAULT_TXD   1024
 #define IXGBEVF_DEFAULT_RXD   512
diff --git a/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c b/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c
index de1ad50..33444b5 100644
--- a/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c
+++ b/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c
@@ -99,6 +99,7 @@ MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
 
 /* forward decls */
 static void ixgbevf_set_itr(struct ixgbevf_q_vector *q_vector);
+static void ixgbevf_free_all_rx_resources(struct ixgbevf_adapter *adapter);
 
 static inline void ixgbevf_release_rx_desc(struct ixgbe_hw *hw,
 					   struct ixgbevf_ring *rx_ring,
@@ -1335,7 +1336,8 @@ static void ixgbevf_init_last_counter_stats(struct ixgbevf_adapter *adapter)
 static void ixgbevf_negotiate_api(struct ixgbevf_adapter *adapter)
 {
 	struct ixgbe_hw *hw = &adapter->hw;
-	int api[] = { ixgbe_mbox_api_10,
+	int api[] = { ixgbe_mbox_api_11,
+		      ixgbe_mbox_api_10,
 		      ixgbe_mbox_api_unknown };
 	int err = 0, idx = 0;
 
@@ -1413,12 +1415,87 @@ static void ixgbevf_up_complete(struct ixgbevf_adapter *adapter)
 	mod_timer(&adapter->watchdog_timer, jiffies);
 }
 
+static int ixgbevf_reset_queues(struct ixgbevf_adapter *adapter)
+{
+	struct ixgbe_hw *hw = &adapter->hw;
+	struct ixgbevf_ring *rx_ring;
+	unsigned int def_q = 0;
+	unsigned int num_tcs = 0;
+	unsigned int num_rx_queues = 1;
+	int err, i;
+
+	spin_lock(&adapter->mbx_lock);
+
+	/* fetch queue configuration from the PF */
+	err = ixgbevf_get_queues(hw, &num_tcs, &def_q);
+
+	spin_unlock(&adapter->mbx_lock);
+
+	if (err)
+		return err;
+
+	if (num_tcs > 1) {
+		/* update default Tx ring register index */
+		adapter->tx_ring[0].reg_idx = def_q;
+
+		/* we need as many queues as traffic classes */
+		num_rx_queues = num_tcs;
+	}
+
+	/* nothing to do if we have the correct number of queues */
+	if (adapter->num_rx_queues == num_rx_queues)
+		return 0;
+
+	/* allocate new rings */
+	rx_ring = kcalloc(num_rx_queues,
+			  sizeof(struct ixgbevf_ring), GFP_KERNEL);
+	if (!rx_ring)
+		return -ENOMEM;
+
+	/* setup ring fields */
+	for (i = 0; i < num_rx_queues; i++) {
+		rx_ring[i].count = adapter->rx_ring_count;
+		rx_ring[i].queue_index = i;
+		rx_ring[i].reg_idx = i;
+		rx_ring[i].dev = &adapter->pdev->dev;
+		rx_ring[i].netdev = adapter->netdev;
+
+		/* allocate resources on the ring */
+		err = ixgbevf_setup_rx_resources(adapter, &rx_ring[i]);
+		if (err) {
+			while (i) {
+				i--;
+				ixgbevf_free_rx_resources(adapter, &rx_ring[i]);
+			}
+			kfree(rx_ring);
+			return err;
+		}
+	}
+
+	/* free the existing rings and queues */
+	ixgbevf_free_all_rx_resources(adapter);
+	adapter->num_rx_queues = 0;
+	kfree(adapter->rx_ring);
+
+	/* move new rings into position on the adapter struct */
+	adapter->rx_ring = rx_ring;
+	adapter->num_rx_queues = num_rx_queues;
+
+	/* reset ring to vector mapping */
+	ixgbevf_reset_q_vectors(adapter);
+	ixgbevf_map_rings_to_vectors(adapter);
+
+	return 0;
+}
+
 void ixgbevf_up(struct ixgbevf_adapter *adapter)
 {
 	struct ixgbe_hw *hw = &adapter->hw;
 
 	ixgbevf_negotiate_api(adapter);
 
+	ixgbevf_reset_queues(adapter);
+
 	ixgbevf_configure(adapter);
 
 	ixgbevf_up_complete(adapter);
@@ -1717,6 +1794,7 @@ static int ixgbevf_alloc_queues(struct ixgbevf_adapter *adapter)
 	for (i = 0; i < adapter->num_tx_queues; i++) {
 		adapter->tx_ring[i].count = adapter->tx_ring_count;
 		adapter->tx_ring[i].queue_index = i;
+		/* reg_idx may be remapped later by DCB config */
 		adapter->tx_ring[i].reg_idx = i;
 		adapter->tx_ring[i].dev = &adapter->pdev->dev;
 		adapter->tx_ring[i].netdev = adapter->netdev;
@@ -1950,8 +2028,11 @@ static int __devinit ixgbevf_sw_init(struct ixgbevf_adapter *adapter)
 	hw->subsystem_device_id = pdev->subsystem_device;
 
 	hw->mbx.ops.init_params(hw);
-	hw->mac.max_tx_queues = MAX_TX_QUEUES;
-	hw->mac.max_rx_queues = MAX_RX_QUEUES;
+
+	/* assume legacy case in which PF would only give VF 2 queues */
+	hw->mac.max_tx_queues = 2;
+	hw->mac.max_rx_queues = 2;
+
 	err = hw->mac.ops.reset_hw(hw);
 	if (err) {
 		dev_info(&pdev->dev,
@@ -2377,6 +2458,63 @@ static void ixgbevf_free_all_rx_resources(struct ixgbevf_adapter *adapter)
 						  &adapter->rx_ring[i]);
 }
 
+static int ixgbevf_setup_queues(struct ixgbevf_adapter *adapter)
+{
+	struct ixgbe_hw *hw = &adapter->hw;
+	struct ixgbevf_ring *rx_ring;
+	unsigned int def_q = 0;
+	unsigned int num_tcs = 0;
+	unsigned int num_rx_queues = 1;
+	int err, i;
+
+	spin_lock(&adapter->mbx_lock);
+
+	/* fetch queue configuration from the PF */
+	err = ixgbevf_get_queues(hw, &num_tcs, &def_q);
+
+	spin_unlock(&adapter->mbx_lock);
+
+	if (err)
+		return err;
+
+	if (num_tcs > 1) {
+		/* update default Tx ring register index */
+		adapter->tx_ring[0].reg_idx = def_q;
+
+		/* we need as many queues as traffic classes */
+		num_rx_queues = num_tcs;
+	}
+
+	/* nothing to do if we have the correct number of queues */
+	if (adapter->num_rx_queues == num_rx_queues)
+		return 0;
+
+	/* allocate new rings */
+	rx_ring = kcalloc(num_rx_queues,
+			  sizeof(struct ixgbevf_ring), GFP_KERNEL);
+	if (!rx_ring)
+		return -ENOMEM;
+
+	/* setup ring fields */
+	for (i = 0; i < num_rx_queues; i++) {
+		rx_ring[i].count = adapter->rx_ring_count;
+		rx_ring[i].queue_index = i;
+		rx_ring[i].reg_idx = i;
+		rx_ring[i].dev = &adapter->pdev->dev;
+		rx_ring[i].netdev = adapter->netdev;
+	}
+
+	/* free the existing ring and queues */
+	adapter->num_rx_queues = 0;
+	kfree(adapter->rx_ring);
+
+	/* move new rings into position on the adapter struct */
+	adapter->rx_ring = rx_ring;
+	adapter->num_rx_queues = num_rx_queues;
+
+	return 0;
+}
+
 /**
  * ixgbevf_open - Called when a network interface is made active
  * @netdev: network interface device structure
@@ -2413,6 +2551,11 @@ static int ixgbevf_open(struct net_device *netdev)
 
 	ixgbevf_negotiate_api(adapter);
 
+	/* setup queue reg_idx and Rx queue count */
+	err = ixgbevf_setup_queues(adapter);
+	if (err)
+		goto err_setup_queues;
+
 	/* allocate transmit descriptors */
 	err = ixgbevf_setup_all_tx_resources(adapter);
 	if (err)
@@ -2451,6 +2594,7 @@ err_setup_rx:
 	ixgbevf_free_all_rx_resources(adapter);
 err_setup_tx:
 	ixgbevf_free_all_tx_resources(adapter);
+err_setup_queues:
 	ixgbevf_reset(adapter);
 
 err_setup_reset:
@@ -2925,8 +3069,15 @@ static int ixgbevf_change_mtu(struct net_device *netdev, int new_mtu)
 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
 	int max_possible_frame = MAXIMUM_ETHERNET_VLAN_SIZE;
 
-	if (adapter->hw.mac.type == ixgbe_mac_X540_vf)
+	switch (adapter->hw.api_version) {
+	case ixgbe_mbox_api_11:
 		max_possible_frame = IXGBE_MAX_JUMBO_FRAME_SIZE;
+		break;
+	default:
+		if (adapter->hw.mac.type == ixgbe_mac_X540_vf)
+			max_possible_frame = IXGBE_MAX_JUMBO_FRAME_SIZE;
+		break;
+	}
 
 	/* MTU < 68 is an error and causes problems on some kernels */
 	if ((new_mtu < 68) || (max_frame > max_possible_frame))
diff --git a/drivers/net/ethernet/intel/ixgbevf/mbx.h b/drivers/net/ethernet/intel/ixgbevf/mbx.h
index 946ce86..0bc3005 100644
--- a/drivers/net/ethernet/intel/ixgbevf/mbx.h
+++ b/drivers/net/ethernet/intel/ixgbevf/mbx.h
@@ -85,6 +85,7 @@
 enum ixgbe_pfvf_api_rev {
 	ixgbe_mbox_api_10,	/* API version 1.0, linux/freebsd VF driver */
 	ixgbe_mbox_api_20,	/* API version 2.0, solaris Phase1 VF driver */
+	ixgbe_mbox_api_11,	/* API version 1.1, linux/freebsd VF driver */
 	/* This value should always be last */
 	ixgbe_mbox_api_unknown,	/* indicates that API version is not known */
 };
@@ -100,6 +101,15 @@ enum ixgbe_pfvf_api_rev {
 #define IXGBE_VF_SET_MACVLAN	0x06 /* VF requests PF for unicast filter */
 #define IXGBE_VF_API_NEGOTIATE	0x08 /* negotiate API version */
 
+/* mailbox API, version 1.1 VF requests */
+#define IXGBE_VF_GET_QUEUE	0x09 /* get queue configuration */
+
+/* GET_QUEUES return data indices within the mailbox */
+#define IXGBE_VF_TX_QUEUES	1	/* number of Tx queues supported */
+#define IXGBE_VF_RX_QUEUES	2	/* number of Rx queues supported */
+#define IXGBE_VF_TRANS_VLAN	3	/* Indication of port vlan */
+#define IXGBE_VF_DEF_QUEUE	4	/* Default queue offset */
+
 /* length of permanent address message returned from PF */
 #define IXGBE_VF_PERMADDR_MSG_LEN 4
 /* word in permanent address message with the current multicast type */
diff --git a/drivers/net/ethernet/intel/ixgbevf/vf.c b/drivers/net/ethernet/intel/ixgbevf/vf.c
index 0c7447e..5fa397b 100644
--- a/drivers/net/ethernet/intel/ixgbevf/vf.c
+++ b/drivers/net/ethernet/intel/ixgbevf/vf.c
@@ -513,6 +513,64 @@ int ixgbevf_negotiate_api_version(struct ixgbe_hw *hw, int api)
 	return err;
 }
 
+int ixgbevf_get_queues(struct ixgbe_hw *hw, unsigned int *num_tcs,
+		       unsigned int *default_tc)
+{
+	int err;
+	u32 msg[5];
+
+	/* do nothing if API doesn't support ixgbevf_get_queues */
+	switch (hw->api_version) {
+	case ixgbe_mbox_api_11:
+		break;
+	default:
+		return 0;
+	}
+
+	/* Fetch queue configuration from the PF */
+	msg[0] = IXGBE_VF_GET_QUEUE;
+	msg[1] = msg[2] = msg[3] = msg[4] = 0;
+	err = hw->mbx.ops.write_posted(hw, msg, 5);
+
+	if (!err)
+		err = hw->mbx.ops.read_posted(hw, msg, 5);
+
+	if (!err) {
+		msg[0] &= ~IXGBE_VT_MSGTYPE_CTS;
+
+		/*
+		 * if we we didn't get an ACK there must have been
+		 * some sort of mailbox error so we should treat it
+		 * as such
+		 */
+		if (msg[0] != (IXGBE_VF_GET_QUEUE | IXGBE_VT_MSGTYPE_ACK))
+			return IXGBE_ERR_MBX;
+
+		/* record and validate values from message */
+		hw->mac.max_tx_queues = msg[IXGBE_VF_TX_QUEUES];
+		if (hw->mac.max_tx_queues == 0 ||
+		    hw->mac.max_tx_queues > IXGBE_VF_MAX_TX_QUEUES)
+			hw->mac.max_tx_queues = IXGBE_VF_MAX_TX_QUEUES;
+
+		hw->mac.max_rx_queues = msg[IXGBE_VF_RX_QUEUES];
+		if (hw->mac.max_rx_queues == 0 ||
+		    hw->mac.max_rx_queues > IXGBE_VF_MAX_RX_QUEUES)
+			hw->mac.max_rx_queues = IXGBE_VF_MAX_RX_QUEUES;
+
+		*num_tcs = msg[IXGBE_VF_TRANS_VLAN];
+		/* in case of unknown state assume we cannot tag frames */
+		if (*num_tcs > hw->mac.max_rx_queues)
+			*num_tcs = 1;
+
+		*default_tc = msg[IXGBE_VF_DEF_QUEUE];
+		/* default to queue 0 on out-of-bounds queue number */
+		if (*default_tc >= hw->mac.max_tx_queues)
+			*default_tc = 0;
+	}
+
+	return err;
+}
+
 static const struct ixgbe_mac_operations ixgbevf_mac_ops = {
 	.init_hw             = ixgbevf_init_hw_vf,
 	.reset_hw            = ixgbevf_reset_hw_vf,
diff --git a/drivers/net/ethernet/intel/ixgbevf/vf.h b/drivers/net/ethernet/intel/ixgbevf/vf.h
index 47f11a5..7b1f502 100644
--- a/drivers/net/ethernet/intel/ixgbevf/vf.h
+++ b/drivers/net/ethernet/intel/ixgbevf/vf.h
@@ -174,5 +174,7 @@ struct ixgbevf_info {
 
 void ixgbevf_rlpml_set_vf(struct ixgbe_hw *hw, u16 max_size);
 int ixgbevf_negotiate_api_version(struct ixgbe_hw *hw, int api);
+int ixgbevf_get_queues(struct ixgbe_hw *hw, unsigned int *num_tcs,
+		       unsigned int *default_tc);
 #endif /* __IXGBE_VF_H__ */
 
-- 
1.7.11.7

[net-next 06/13] ixgbe: add WOL support for new subdevice id

[Jeff Kirsher]

From: Emil Tantilov <emil.s.tantilov@intel.com>

This patch adds a subdevice id for new 82599 device. The define is needed
to allow enabling WOL support.

Signed-off-by: Emil Tantilov <emil.s.tantilov@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 1 +
 drivers/net/ethernet/intel/ixgbe/ixgbe_type.h | 1 +
 2 files changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index 35be7d3..e417394 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -7080,6 +7080,7 @@ int ixgbe_wol_supported(struct ixgbe_adapter *adapter, u16 device_id,
 				break;
 		case IXGBE_SUBDEV_ID_82599_SFP:
 		case IXGBE_SUBDEV_ID_82599_RNDC:
+		case IXGBE_SUBDEV_ID_82599_ECNA_DP:
 			is_wol_supported = 1;
 			break;
 		}
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h b/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h
index 0722f33..21915e2 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_type.h
@@ -56,6 +56,7 @@
 #define IXGBE_SUBDEV_ID_82599_SFP        0x11A9
 #define IXGBE_SUBDEV_ID_82599_RNDC       0x1F72
 #define IXGBE_SUBDEV_ID_82599_560FLR     0x17D0
+#define IXGBE_SUBDEV_ID_82599_ECNA_DP    0x0470
 #define IXGBE_DEV_ID_82599_SFP_EM        0x1507
 #define IXGBE_DEV_ID_82599_SFP_SF2       0x154D
 #define IXGBE_DEV_ID_82599EN_SFP         0x1557
-- 
1.7.11.7

[net-next 07/13] ixgbe: (PTP) refactor init, cyclecounter and reset

[Jeff Kirsher]

From: Jacob Keller <jacob.e.keller@intel.com>

This patch modifies when and where PTP registers and data are set. Previously
a work-around was used inside cyclecounter_start in order to reset some of the
time registers. This patch creates a new ixgbe_ptp_reset specifically for this
purpose. The cyclecounter configuration has trimmed down to only modify what
is necessary. Due to hardware conditions after probe and before open, PTP init
has now moved into the ixgbe_open call. This allows the ptp device name in the
sysfs to be the ethernet device name instead of the MAC address.

The cyclecounter check flag is renamed to PTP_ENABLED and is used to prevent
PTP init from happening when PTP has not been enabled.

CC: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h      |   4 +-
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c |  26 +++---
 drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c  | 109 +++++++++++---------------
 3 files changed, 66 insertions(+), 73 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
index 101e525..c64a777 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
@@ -483,7 +483,7 @@ struct ixgbe_adapter {
 #define IXGBE_FLAG2_FDIR_REQUIRES_REINIT        (u32)(1 << 7)
 #define IXGBE_FLAG2_RSS_FIELD_IPV4_UDP		(u32)(1 << 8)
 #define IXGBE_FLAG2_RSS_FIELD_IPV6_UDP		(u32)(1 << 9)
-#define IXGBE_FLAG2_OVERFLOW_CHECK_ENABLED	(u32)(1 << 10)
+#define IXGBE_FLAG2_PTP_ENABLED			(u32)(1 << 10)
 #define IXGBE_FLAG2_PTP_PPS_ENABLED		(u32)(1 << 11)
 
 	/* Tx fast path data */
@@ -581,7 +581,6 @@ struct ixgbe_adapter {
 	struct timecounter tc;
 	int rx_hwtstamp_filter;
 	u32 base_incval;
-	u32 cycle_speed;
 #endif /* CONFIG_IXGBE_PTP */
 
 	/* SR-IOV */
@@ -754,6 +753,7 @@ extern void ixgbe_ptp_rx_hwtstamp(struct ixgbe_q_vector *q_vector,
 extern int ixgbe_ptp_hwtstamp_ioctl(struct ixgbe_adapter *adapter,
 				    struct ifreq *ifr, int cmd);
 extern void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter);
+extern void ixgbe_ptp_reset(struct ixgbe_adapter *adapter);
 extern void ixgbe_ptp_check_pps_event(struct ixgbe_adapter *adapter, u32 eicr);
 #endif /* CONFIG_IXGBE_PTP */
 
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index e417394..46bf0fd 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -4210,6 +4210,11 @@ void ixgbe_reset(struct ixgbe_adapter *adapter)
 	/* update SAN MAC vmdq pool selection */
 	if (hw->mac.san_mac_rar_index)
 		hw->mac.ops.set_vmdq_san_mac(hw, VMDQ_P(0));
+
+#ifdef CONFIG_IXGBE_PTP
+	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
+		ixgbe_ptp_reset(adapter);
+#endif
 }
 
 /**
@@ -4919,6 +4924,10 @@ static int ixgbe_open(struct net_device *netdev)
 	if (err)
 		goto err_set_queues;
 
+#ifdef CONFIG_IXGBE_PTP
+	ixgbe_ptp_init(adapter);
+#endif /* CONFIG_IXGBE_PTP*/
+
 	ixgbe_up_complete(adapter);
 
 	return 0;
@@ -4950,6 +4959,10 @@ static int ixgbe_close(struct net_device *netdev)
 {
 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
 
+#ifdef CONFIG_IXGBE_PTP
+	ixgbe_ptp_stop(adapter);
+#endif
+
 	ixgbe_down(adapter);
 	ixgbe_free_irq(adapter);
 
@@ -5518,7 +5531,8 @@ static void ixgbe_watchdog_link_is_up(struct ixgbe_adapter *adapter)
 	}
 
 #ifdef CONFIG_IXGBE_PTP
-	ixgbe_ptp_start_cyclecounter(adapter);
+	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
+		ixgbe_ptp_start_cyclecounter(adapter);
 #endif
 
 	e_info(drv, "NIC Link is Up %s, Flow Control: %s\n",
@@ -5565,7 +5579,8 @@ static void ixgbe_watchdog_link_is_down(struct ixgbe_adapter *adapter)
 		adapter->flags2 |= IXGBE_FLAG2_SEARCH_FOR_SFP;
 
 #ifdef CONFIG_IXGBE_PTP
-	ixgbe_ptp_start_cyclecounter(adapter);
+	if (adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED)
+		ixgbe_ptp_start_cyclecounter(adapter);
 #endif
 
 	e_info(drv, "NIC Link is Down\n");
@@ -7403,10 +7418,6 @@ static int __devinit ixgbe_probe(struct pci_dev *pdev,
 
 	device_set_wakeup_enable(&adapter->pdev->dev, adapter->wol);
 
-#ifdef CONFIG_IXGBE_PTP
-	ixgbe_ptp_init(adapter);
-#endif /* CONFIG_IXGBE_PTP*/
-
 	/* save off EEPROM version number */
 	hw->eeprom.ops.read(hw, 0x2e, &adapter->eeprom_verh);
 	hw->eeprom.ops.read(hw, 0x2d, &adapter->eeprom_verl);
@@ -7544,9 +7555,6 @@ static void __devexit ixgbe_remove(struct pci_dev *pdev)
 	set_bit(__IXGBE_DOWN, &adapter->state);
 	cancel_work_sync(&adapter->service_task);
 
-#ifdef CONFIG_IXGBE_PTP
-	ixgbe_ptp_stop(adapter);
-#endif
 
 #ifdef CONFIG_IXGBE_DCA
 	if (adapter->flags & IXGBE_FLAG_DCA_ENABLED) {
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
index d929131..5e71ddb 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ptp.c
@@ -411,7 +411,7 @@ void ixgbe_ptp_overflow_check(struct ixgbe_adapter *adapter)
 	unsigned long elapsed_jiffies = adapter->last_overflow_check - jiffies;
 	struct timespec ts;
 
-	if ((adapter->flags2 & IXGBE_FLAG2_OVERFLOW_CHECK_ENABLED) &&
+	if ((adapter->flags2 & IXGBE_FLAG2_PTP_ENABLED) &&
 	    (elapsed_jiffies >= IXGBE_OVERFLOW_PERIOD)) {
 		ixgbe_ptp_gettime(&adapter->ptp_caps, &ts);
 		adapter->last_overflow_check = jiffies;
@@ -759,58 +759,20 @@ int ixgbe_ptp_hwtstamp_ioctl(struct ixgbe_adapter *adapter,
  * ixgbe_ptp_start_cyclecounter - create the cycle counter from hw
  * @adapter: pointer to the adapter structure
  *
- * this function initializes the timecounter and cyclecounter
- * structures for use in generated a ns counter from the arbitrary
- * fixed point cycles registers in the hardware.
- *
- * A change in link speed impacts the frequency of the DMA clock on
- * the device, which is used to generate the cycle counter
- * registers. Therefor this function is called whenever the link speed
- * changes.
- *
- * This function also turns on the SDP pin for clock out feature (X540
- * only), because this is where the shift is first calculated.
+ * This function should be called to set the proper values for the TIMINCA
+ * register and tell the cyclecounter structure what the tick rate of SYSTIME
+ * is. It does not directly modify SYSTIME registers or the timecounter
+ * structure. It should be called whenever a new TIMINCA value is necessary,
+ * such as during initialization or when the link speed changes.
  */
 void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter)
 {
 	struct ixgbe_hw *hw = &adapter->hw;
 	u32 incval = 0;
-	u32 timinca = 0;
 	u32 shift = 0;
-	u32 cycle_speed;
 	unsigned long flags;
 
 	/**
-	 * Determine what speed we need to set the cyclecounter
-	 * for. It should be different for 100Mb, 1Gb, and 10Gb. Treat
-	 * unknown speeds as 10Gb. (Hence why we can't just copy the
-	 * link_speed.
-	 */
-	switch (adapter->link_speed) {
-	case IXGBE_LINK_SPEED_100_FULL:
-	case IXGBE_LINK_SPEED_1GB_FULL:
-	case IXGBE_LINK_SPEED_10GB_FULL:
-		cycle_speed = adapter->link_speed;
-		break;
-	default:
-		/* cycle speed should be 10Gb when there is no link */
-		cycle_speed = IXGBE_LINK_SPEED_10GB_FULL;
-		break;
-	}
-
-	/*
-	 * grab the current TIMINCA value from the register so that it can be
-	 * double checked. If the register value has been cleared, it must be
-	 * reset to the correct value for generating a cyclecounter. If
-	 * TIMINCA is zero, the SYSTIME registers do not increment at all.
-	 */
-	timinca = IXGBE_READ_REG(hw, IXGBE_TIMINCA);
-
-	/* Bail if the cycle speed didn't change and TIMINCA is non-zero */
-	if (adapter->cycle_speed == cycle_speed && timinca)
-		return;
-
-	/**
 	 * Scale the NIC cycle counter by a large factor so that
 	 * relatively small corrections to the frequency can be added
 	 * or subtracted. The drawbacks of a large factor include
@@ -819,8 +781,12 @@ void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter)
 	 * to nanoseconds using only a multiplier and a right-shift,
 	 * and (c) the value must fit within the timinca register space
 	 * => math based on internal DMA clock rate and available bits
+	 *
+	 * Note that when there is no link, internal DMA clock is same as when
+	 * link speed is 10Gb. Set the registers correctly even when link is
+	 * down to preserve the clock setting
 	 */
-	switch (cycle_speed) {
+	switch (adapter->link_speed) {
 	case IXGBE_LINK_SPEED_100_FULL:
 		incval = IXGBE_INCVAL_100;
 		shift = IXGBE_INCVAL_SHIFT_100;
@@ -830,6 +796,7 @@ void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter)
 		shift = IXGBE_INCVAL_SHIFT_1GB;
 		break;
 	case IXGBE_LINK_SPEED_10GB_FULL:
+	default:
 		incval = IXGBE_INCVAL_10GB;
 		shift = IXGBE_INCVAL_SHIFT_10GB;
 		break;
@@ -857,18 +824,11 @@ void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter)
 		return;
 	}
 
-	/* reset the system time registers */
-	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x00000000);
-	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x00000000);
-	IXGBE_WRITE_FLUSH(hw);
-
-	/* store the new cycle speed */
-	adapter->cycle_speed = cycle_speed;
-
+	/* update the base incval used to calculate frequency adjustment */
 	ACCESS_ONCE(adapter->base_incval) = incval;
 	smp_mb();
 
-	/* grab the ptp lock */
+	/* need lock to prevent incorrect read while modifying cyclecounter */
 	spin_lock_irqsave(&adapter->tmreg_lock, flags);
 
 	memset(&adapter->cc, 0, sizeof(adapter->cc));
@@ -877,6 +837,31 @@ void ixgbe_ptp_start_cyclecounter(struct ixgbe_adapter *adapter)
 	adapter->cc.shift = shift;
 	adapter->cc.mult = 1;
 
+	spin_unlock_irqrestore(&adapter->tmreg_lock, flags);
+}
+
+/**
+ * ixgbe_ptp_reset
+ * @adapter: the ixgbe private board structure
+ *
+ * When the MAC resets, all timesync features are reset. This function should be
+ * called to re-enable the PTP clock structure. It will re-init the timecounter
+ * structure based on the kernel time as well as setup the cycle counter data.
+ */
+void ixgbe_ptp_reset(struct ixgbe_adapter *adapter)
+{
+	struct ixgbe_hw *hw = &adapter->hw;
+	unsigned long flags;
+
+	/* set SYSTIME registers to 0 just in case */
+	IXGBE_WRITE_REG(hw, IXGBE_SYSTIML, 0x00000000);
+	IXGBE_WRITE_REG(hw, IXGBE_SYSTIMH, 0x00000000);
+	IXGBE_WRITE_FLUSH(hw);
+
+	ixgbe_ptp_start_cyclecounter(adapter);
+
+	spin_lock_irqsave(&adapter->tmreg_lock, flags);
+
 	/* reset the ns time counter */
 	timecounter_init(&adapter->tc, &adapter->cc,
 			 ktime_to_ns(ktime_get_real()));
@@ -904,7 +889,7 @@ void ixgbe_ptp_init(struct ixgbe_adapter *adapter)
 
 	switch (adapter->hw.mac.type) {
 	case ixgbe_mac_X540:
-		snprintf(adapter->ptp_caps.name, 16, "%pm", netdev->dev_addr);
+		snprintf(adapter->ptp_caps.name, 16, "%s", netdev->name);
 		adapter->ptp_caps.owner = THIS_MODULE;
 		adapter->ptp_caps.max_adj = 250000000;
 		adapter->ptp_caps.n_alarm = 0;
@@ -918,7 +903,7 @@ void ixgbe_ptp_init(struct ixgbe_adapter *adapter)
 		adapter->ptp_caps.enable = ixgbe_ptp_enable;
 		break;
 	case ixgbe_mac_82599EB:
-		snprintf(adapter->ptp_caps.name, 16, "%pm", netdev->dev_addr);
+		snprintf(adapter->ptp_caps.name, 16, "%s", netdev->name);
 		adapter->ptp_caps.owner = THIS_MODULE;
 		adapter->ptp_caps.max_adj = 250000000;
 		adapter->ptp_caps.n_alarm = 0;
@@ -942,11 +927,6 @@ void ixgbe_ptp_init(struct ixgbe_adapter *adapter)
 
 	spin_lock_init(&adapter->tmreg_lock);
 
-	ixgbe_ptp_start_cyclecounter(adapter);
-
-	/* (Re)start the overflow check */
-	adapter->flags2 |= IXGBE_FLAG2_OVERFLOW_CHECK_ENABLED;
-
 	adapter->ptp_clock = ptp_clock_register(&adapter->ptp_caps,
 						&adapter->pdev->dev);
 	if (IS_ERR(adapter->ptp_clock)) {
@@ -955,6 +935,11 @@ void ixgbe_ptp_init(struct ixgbe_adapter *adapter)
 	} else
 		e_dev_info("registered PHC device on %s\n", netdev->name);
 
+	ixgbe_ptp_reset(adapter);
+
+	/* set the flag that PTP has been enabled */
+	adapter->flags2 |= IXGBE_FLAG2_PTP_ENABLED;
+
 	return;
 }
 
@@ -967,7 +952,7 @@ void ixgbe_ptp_init(struct ixgbe_adapter *adapter)
 void ixgbe_ptp_stop(struct ixgbe_adapter *adapter)
 {
 	/* stop the overflow check task */
-	adapter->flags2 &= ~(IXGBE_FLAG2_OVERFLOW_CHECK_ENABLED |
+	adapter->flags2 &= ~(IXGBE_FLAG2_PTP_ENABLED |
 			     IXGBE_FLAG2_PTP_PPS_ENABLED);
 
 	ixgbe_ptp_setup_sdp(adapter);
-- 
1.7.11.7

[net-next 08/13] ixgbe: using is_zero_ether_addr() to simplify the code

[Jeff Kirsher]

From: Wei Yongjun <yongjun_wei@trendmicro.com.cn>

Using is_zero_ether_addr() to simplify the code.

Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_common.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
index dbf37e4..a2a9bcc 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
@@ -1778,8 +1778,7 @@ s32 ixgbe_validate_mac_addr(u8 *mac_addr)
 	else if (IXGBE_IS_BROADCAST(mac_addr))
 		status = IXGBE_ERR_INVALID_MAC_ADDR;
 	/* Reject the zero address */
-	else if (mac_addr[0] == 0 && mac_addr[1] == 0 && mac_addr[2] == 0 &&
-	         mac_addr[3] == 0 && mac_addr[4] == 0 && mac_addr[5] == 0)
+	else if (is_zero_ether_addr(mac_addr))
 		status = IXGBE_ERR_INVALID_MAC_ADDR;
 
 	return status;
-- 
1.7.11.7

[net-next 09/13] ixgbe: Correcting small packet padding

[Jeff Kirsher]

From: Tushar Dave <tushar.n.dave@intel.com>

Driver pad skb up to 17 bytes because of the HW requirement. However, that code
implementation mess up the skb tail pointer after padding. This patch sets
skb->tail correctly.

Signed-off-by: Tushar Dave <tushar.n.dave@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index 46bf0fd..c1ce6e9 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -6538,6 +6538,7 @@ static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
 		if (skb_pad(skb, 17 - skb->len))
 			return NETDEV_TX_OK;
 		skb->len = 17;
+		skb_set_tail_pointer(skb, 17);
 	}
 
 	tx_ring = adapter->tx_ring[skb->queue_mapping];
-- 
1.7.11.7

[net-next 10/13] ixgbe: Drop unnecessary addition from ixgbe_set_rx_buffer_len

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

We still had some code floating around from the old single buffer receive
path.  As a result we were adding VLAN_HLEN to max_frame although the
resultant value was never used.  Since that is the case we can drop this from
the function.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
index c1ce6e9..88d636a 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_main.c
@@ -3289,9 +3289,6 @@ static void ixgbe_set_rx_buffer_len(struct ixgbe_adapter *adapter)
 		IXGBE_WRITE_REG(hw, IXGBE_MHADD, mhadd);
 	}
 
-	/* MHADD will allow an extra 4 bytes past for vlan tagged frames */
-	max_frame += VLAN_HLEN;
-
 	hlreg0 = IXGBE_READ_REG(hw, IXGBE_HLREG0);
 	/* set jumbo enable since MHADD.MFS is keeping size locked at max_frame */
 	hlreg0 |= IXGBE_HLREG0_JUMBOEN;
-- 
1.7.11.7

[net-next 11/13] ixgbe: Add function ixgbe_reset_pipeline_82599

[Jeff Kirsher]

From: Don Skidmore <donald.c.skidmore@intel.com>

This patch adds a function that forces a full pipeline reset.  This
function will be used in following patches to completely reset the PHY
during resets.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c  | 44 +++++++++++++++++++++++++
 drivers/net/ethernet/intel/ixgbe/ixgbe_common.h |  1 +
 2 files changed, 45 insertions(+)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c
index 1077cb2..b527813 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c
@@ -2080,6 +2080,50 @@ static s32 ixgbe_read_eeprom_82599(struct ixgbe_hw *hw,
 	return ret_val;
 }
 
+/**
+ * ixgbe_reset_pipeline_82599 - perform pipeline reset
+ *
+ * @hw: pointer to hardware structure
+ *
+ * Reset pipeline by asserting Restart_AN together with LMS change to ensure
+ * full pipeline reset.  Note - We must hold the SW/FW semaphore before writing
+ * to AUTOC, so this function assumes the semaphore is held.
+ **/
+s32 ixgbe_reset_pipeline_82599(struct ixgbe_hw *hw)
+{
+	s32 i, autoc_reg, ret_val;
+	s32 anlp1_reg = 0;
+
+	autoc_reg = IXGBE_READ_REG(hw, IXGBE_AUTOC);
+	autoc_reg |= IXGBE_AUTOC_AN_RESTART;
+
+	/* Write AUTOC register with toggled LMS[2] bit and Restart_AN */
+	IXGBE_WRITE_REG(hw, IXGBE_AUTOC, autoc_reg ^ IXGBE_AUTOC_LMS_1G_AN);
+
+	/* Wait for AN to leave state 0 */
+	for (i = 0; i < 10; i++) {
+		usleep_range(4000, 8000);
+		anlp1_reg = IXGBE_READ_REG(hw, IXGBE_ANLP1);
+		if (anlp1_reg & IXGBE_ANLP1_AN_STATE_MASK)
+			break;
+	}
+
+	if (!(anlp1_reg & IXGBE_ANLP1_AN_STATE_MASK)) {
+		hw_dbg(hw, "auto negotiation not completed\n");
+		ret_val = IXGBE_ERR_RESET_FAILED;
+		goto reset_pipeline_out;
+	}
+
+	ret_val = 0;
+
+reset_pipeline_out:
+	/* Write AUTOC register with original LMS field and Restart_AN */
+	IXGBE_WRITE_REG(hw, IXGBE_AUTOC, autoc_reg);
+	IXGBE_WRITE_FLUSH(hw);
+
+	return ret_val;
+}
+
 static struct ixgbe_mac_operations mac_ops_82599 = {
 	.init_hw                = &ixgbe_init_hw_generic,
 	.reset_hw               = &ixgbe_reset_hw_82599,
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.h b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.h
index d813d11..587db47 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.h
@@ -107,6 +107,7 @@ void ixgbe_clear_tx_pending(struct ixgbe_hw *hw);
 
 void ixgbe_set_rxpba_generic(struct ixgbe_hw *hw, int num_pb,
 			     u32 headroom, int strategy);
+s32 ixgbe_reset_pipeline_82599(struct ixgbe_hw *hw);
 
 #define IXGBE_I2C_THERMAL_SENSOR_ADDR	0xF8
 #define IXGBE_EMC_INTERNAL_DATA		0x00
-- 
1.7.11.7

[net-next 12/13] ixgbe: Add support for pipeline reset

[Jeff Kirsher]

From: Don Skidmore <donald.c.skidmore@intel.com>

Calling the ixgbe_reset_pipeline_82599 function will ensure a full pipeline
reset on all 82599 devices.  This is necessary to avoid possible link issues.
Since this patch accomplishes this by modifying AUTOC.LMS we need to wrap
all AUTOC writes when LESM is enabled.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe.h        |   1 +
 drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c  | 105 ++++++++++++++++++------
 drivers/net/ethernet/intel/ixgbe/ixgbe_common.c |  70 +++++++++++++++-
 3 files changed, 149 insertions(+), 27 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe.h b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
index c64a777..a17116b 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe.h
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe.h
@@ -693,6 +693,7 @@ extern s32 ixgbe_fdir_erase_perfect_filter_82599(struct ixgbe_hw *hw,
 						 u16 soft_id);
 extern void ixgbe_atr_compute_perfect_hash_82599(union ixgbe_atr_input *input,
 						 union ixgbe_atr_input *mask);
+extern bool ixgbe_verify_lesm_fw_enabled_82599(struct ixgbe_hw *hw);
 extern void ixgbe_set_rx_mode(struct net_device *netdev);
 #ifdef CONFIG_IXGBE_DCB
 extern void ixgbe_set_rx_drop_en(struct ixgbe_adapter *adapter);
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c
index b527813..3e49fcb 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_82599.c
@@ -62,7 +62,6 @@ static s32 ixgbe_setup_copper_link_82599(struct ixgbe_hw *hw,
                                          bool autoneg,
                                          bool autoneg_wait_to_complete);
 static s32 ixgbe_verify_fw_version_82599(struct ixgbe_hw *hw);
-static bool ixgbe_verify_lesm_fw_enabled_82599(struct ixgbe_hw *hw);
 
 static void ixgbe_init_mac_link_ops_82599(struct ixgbe_hw *hw)
 {
@@ -99,9 +98,8 @@ static void ixgbe_init_mac_link_ops_82599(struct ixgbe_hw *hw)
 static s32 ixgbe_setup_sfp_modules_82599(struct ixgbe_hw *hw)
 {
 	s32 ret_val = 0;
-	u32 reg_anlp1 = 0;
-	u32 i = 0;
 	u16 list_offset, data_offset, data_value;
+	bool got_lock = false;
 
 	if (hw->phy.sfp_type != ixgbe_sfp_type_unknown) {
 		ixgbe_init_mac_link_ops_82599(hw);
@@ -137,28 +135,37 @@ static s32 ixgbe_setup_sfp_modules_82599(struct ixgbe_hw *hw)
 		usleep_range(hw->eeprom.semaphore_delay * 1000,
 			     hw->eeprom.semaphore_delay * 2000);
 
+		/* Need SW/FW semaphore around AUTOC writes if LESM on,
+		 * likewise reset_pipeline requires lock as it also writes
+		 * AUTOC.
+		 */
+		if (ixgbe_verify_lesm_fw_enabled_82599(hw)) {
+			ret_val = hw->mac.ops.acquire_swfw_sync(hw,
+							IXGBE_GSSR_MAC_CSR_SM);
+			if (ret_val)
+				goto setup_sfp_out;
+
+			got_lock = true;
+		}
+
 		/* Now restart DSP by setting Restart_AN and clearing LMS */
 		IXGBE_WRITE_REG(hw, IXGBE_AUTOC, ((IXGBE_READ_REG(hw,
 		                IXGBE_AUTOC) & ~IXGBE_AUTOC_LMS_MASK) |
 		                IXGBE_AUTOC_AN_RESTART));
 
-		/* Wait for AN to leave state 0 */
-		for (i = 0; i < 10; i++) {
-			usleep_range(4000, 8000);
-			reg_anlp1 = IXGBE_READ_REG(hw, IXGBE_ANLP1);
-			if (reg_anlp1 & IXGBE_ANLP1_AN_STATE_MASK)
-				break;
+		ret_val = ixgbe_reset_pipeline_82599(hw);
+
+		if (got_lock) {
+			hw->mac.ops.release_swfw_sync(hw,
+						      IXGBE_GSSR_MAC_CSR_SM);
+			got_lock = false;
 		}
-		if (!(reg_anlp1 & IXGBE_ANLP1_AN_STATE_MASK)) {
-			hw_dbg(hw, "sfp module setup not complete\n");
+
+		if (ret_val) {
+			hw_dbg(hw, " sfp module setup not complete\n");
 			ret_val = IXGBE_ERR_SFP_SETUP_NOT_COMPLETE;
 			goto setup_sfp_out;
 		}
-
-		/* Restart DSP by setting Restart_AN and return to SFI mode */
-		IXGBE_WRITE_REG(hw, IXGBE_AUTOC, (IXGBE_READ_REG(hw,
-		                IXGBE_AUTOC) | IXGBE_AUTOC_LMS_10G_SERIAL |
-		                IXGBE_AUTOC_AN_RESTART));
 	}
 
 setup_sfp_out:
@@ -394,14 +401,26 @@ static s32 ixgbe_start_mac_link_82599(struct ixgbe_hw *hw,
 	u32 links_reg;
 	u32 i;
 	s32 status = 0;
+	bool got_lock = false;
+
+	if (ixgbe_verify_lesm_fw_enabled_82599(hw)) {
+		status = hw->mac.ops.acquire_swfw_sync(hw,
+						IXGBE_GSSR_MAC_CSR_SM);
+		if (status)
+			goto out;
+
+		got_lock = true;
+	}
 
 	/* Restart link */
-	autoc_reg = IXGBE_READ_REG(hw, IXGBE_AUTOC);
-	autoc_reg |= IXGBE_AUTOC_AN_RESTART;
-	IXGBE_WRITE_REG(hw, IXGBE_AUTOC, autoc_reg);
+	ixgbe_reset_pipeline_82599(hw);
+
+	if (got_lock)
+		hw->mac.ops.release_swfw_sync(hw, IXGBE_GSSR_MAC_CSR_SM);
 
 	/* Only poll for autoneg to complete if specified to do so */
 	if (autoneg_wait_to_complete) {
+		autoc_reg = IXGBE_READ_REG(hw, IXGBE_AUTOC);
 		if ((autoc_reg & IXGBE_AUTOC_LMS_MASK) ==
 		     IXGBE_AUTOC_LMS_KX4_KX_KR ||
 		    (autoc_reg & IXGBE_AUTOC_LMS_MASK) ==
@@ -425,6 +444,7 @@ static s32 ixgbe_start_mac_link_82599(struct ixgbe_hw *hw,
 	/* Add delay to filter out noises during initial link setup */
 	msleep(50);
 
+out:
 	return status;
 }
 
@@ -779,6 +799,7 @@ static s32 ixgbe_setup_mac_link_82599(struct ixgbe_hw *hw,
 	u32 links_reg;
 	u32 i;
 	ixgbe_link_speed link_capabilities = IXGBE_LINK_SPEED_UNKNOWN;
+	bool got_lock = false;
 
 	/* Check to see if speed passed in is supported. */
 	status = hw->mac.ops.get_link_capabilities(hw, &link_capabilities,
@@ -836,9 +857,26 @@ static s32 ixgbe_setup_mac_link_82599(struct ixgbe_hw *hw,
 	}
 
 	if (autoc != start_autoc) {
+		/* Need SW/FW semaphore around AUTOC writes if LESM is on,
+		 * likewise reset_pipeline requires us to hold this lock as
+		 * it also writes to AUTOC.
+		 */
+		if (ixgbe_verify_lesm_fw_enabled_82599(hw)) {
+			status = hw->mac.ops.acquire_swfw_sync(hw,
+							IXGBE_GSSR_MAC_CSR_SM);
+			if (status != 0)
+				goto out;
+
+			got_lock = true;
+		}
+
 		/* Restart link */
-		autoc |= IXGBE_AUTOC_AN_RESTART;
 		IXGBE_WRITE_REG(hw, IXGBE_AUTOC, autoc);
+		ixgbe_reset_pipeline_82599(hw);
+
+		if (got_lock)
+			hw->mac.ops.release_swfw_sync(hw,
+						      IXGBE_GSSR_MAC_CSR_SM);
 
 		/* Only poll for autoneg to complete if specified to do so */
 		if (autoneg_wait_to_complete) {
@@ -994,9 +1032,28 @@ mac_reset_top:
 		hw->mac.orig_autoc2 = autoc2;
 		hw->mac.orig_link_settings_stored = true;
 	} else {
-		if (autoc != hw->mac.orig_autoc)
-			IXGBE_WRITE_REG(hw, IXGBE_AUTOC, (hw->mac.orig_autoc |
-			                IXGBE_AUTOC_AN_RESTART));
+		if (autoc != hw->mac.orig_autoc) {
+			/* Need SW/FW semaphore around AUTOC writes if LESM is
+			 * on, likewise reset_pipeline requires us to hold
+			 * this lock as it also writes to AUTOC.
+			 */
+			bool got_lock = false;
+			if (ixgbe_verify_lesm_fw_enabled_82599(hw)) {
+				status = hw->mac.ops.acquire_swfw_sync(hw,
+							IXGBE_GSSR_MAC_CSR_SM);
+				if (status)
+					goto reset_hw_out;
+
+				got_lock = true;
+			}
+
+			IXGBE_WRITE_REG(hw, IXGBE_AUTOC, hw->mac.orig_autoc);
+			ixgbe_reset_pipeline_82599(hw);
+
+			if (got_lock)
+				hw->mac.ops.release_swfw_sync(hw,
+							IXGBE_GSSR_MAC_CSR_SM);
+		}
 
 		if ((autoc2 & IXGBE_AUTOC2_UPPER_MASK) !=
 		    (hw->mac.orig_autoc2 & IXGBE_AUTOC2_UPPER_MASK)) {
@@ -1983,7 +2040,7 @@ fw_version_out:
  *  Returns true if the LESM FW module is present and enabled. Otherwise
  *  returns false. Smart Speed must be disabled if LESM FW module is enabled.
  **/
-static bool ixgbe_verify_lesm_fw_enabled_82599(struct ixgbe_hw *hw)
+bool ixgbe_verify_lesm_fw_enabled_82599(struct ixgbe_hw *hw)
 {
 	bool lesm_enabled = false;
 	u16 fw_offset, fw_lesm_param_offset, fw_lesm_state;
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
index a2a9bcc..8f285ed 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_common.c
@@ -90,6 +90,7 @@ static s32 ixgbe_setup_fc(struct ixgbe_hw *hw)
 	s32 ret_val = 0;
 	u32 reg = 0, reg_bp = 0;
 	u16 reg_cu = 0;
+	bool got_lock = false;
 
 	/*
 	 * Validate the requested mode.  Strict IEEE mode does not allow
@@ -210,8 +211,29 @@ static s32 ixgbe_setup_fc(struct ixgbe_hw *hw)
 	 *
 	 */
 	if (hw->phy.media_type == ixgbe_media_type_backplane) {
-		reg_bp |= IXGBE_AUTOC_AN_RESTART;
+		/* Need the SW/FW semaphore around AUTOC writes if 82599 and
+		 * LESM is on, likewise reset_pipeline requries the lock as
+		 * it also writes AUTOC.
+		 */
+		if ((hw->mac.type == ixgbe_mac_82599EB) &&
+		    ixgbe_verify_lesm_fw_enabled_82599(hw)) {
+			ret_val = hw->mac.ops.acquire_swfw_sync(hw,
+							IXGBE_GSSR_MAC_CSR_SM);
+			if (ret_val)
+				goto out;
+
+			got_lock = true;
+		}
+
 		IXGBE_WRITE_REG(hw, IXGBE_AUTOC, reg_bp);
+
+		if (hw->mac.type == ixgbe_mac_82599EB)
+			ixgbe_reset_pipeline_82599(hw);
+
+		if (got_lock)
+			hw->mac.ops.release_swfw_sync(hw,
+						      IXGBE_GSSR_MAC_CSR_SM);
+
 	} else if ((hw->phy.media_type == ixgbe_media_type_copper) &&
 		    (ixgbe_device_supports_autoneg_fc(hw) == 0)) {
 		hw->phy.ops.write_reg(hw, MDIO_AN_ADVERTISE,
@@ -2616,6 +2638,7 @@ s32 ixgbe_blink_led_start_generic(struct ixgbe_hw *hw, u32 index)
 	bool link_up = false;
 	u32 autoc_reg = IXGBE_READ_REG(hw, IXGBE_AUTOC);
 	u32 led_reg = IXGBE_READ_REG(hw, IXGBE_LEDCTL);
+	s32 ret_val = 0;
 
 	/*
 	 * Link must be up to auto-blink the LEDs;
@@ -2624,10 +2647,28 @@ s32 ixgbe_blink_led_start_generic(struct ixgbe_hw *hw, u32 index)
 	hw->mac.ops.check_link(hw, &speed, &link_up, false);
 
 	if (!link_up) {
+		/* Need the SW/FW semaphore around AUTOC writes if 82599 and
+		 * LESM is on.
+		 */
+		bool got_lock = false;
+
+		if ((hw->mac.type == ixgbe_mac_82599EB) &&
+		    ixgbe_verify_lesm_fw_enabled_82599(hw)) {
+			ret_val = hw->mac.ops.acquire_swfw_sync(hw,
+							IXGBE_GSSR_MAC_CSR_SM);
+			if (ret_val)
+				goto out;
+
+			got_lock = true;
+		}
 		autoc_reg |= IXGBE_AUTOC_AN_RESTART;
 		autoc_reg |= IXGBE_AUTOC_FLU;
 		IXGBE_WRITE_REG(hw, IXGBE_AUTOC, autoc_reg);
 		IXGBE_WRITE_FLUSH(hw);
+
+		if (got_lock)
+			hw->mac.ops.release_swfw_sync(hw,
+						      IXGBE_GSSR_MAC_CSR_SM);
 		usleep_range(10000, 20000);
 	}
 
@@ -2636,7 +2677,8 @@ s32 ixgbe_blink_led_start_generic(struct ixgbe_hw *hw, u32 index)
 	IXGBE_WRITE_REG(hw, IXGBE_LEDCTL, led_reg);
 	IXGBE_WRITE_FLUSH(hw);
 
-	return 0;
+out:
+	return ret_val;
 }
 
 /**
@@ -2648,18 +2690,40 @@ s32 ixgbe_blink_led_stop_generic(struct ixgbe_hw *hw, u32 index)
 {
 	u32 autoc_reg = IXGBE_READ_REG(hw, IXGBE_AUTOC);
 	u32 led_reg = IXGBE_READ_REG(hw, IXGBE_LEDCTL);
+	s32 ret_val = 0;
+	bool got_lock = false;
+
+	/* Need the SW/FW semaphore around AUTOC writes if 82599 and
+	 * LESM is on.
+	 */
+	if ((hw->mac.type == ixgbe_mac_82599EB) &&
+	    ixgbe_verify_lesm_fw_enabled_82599(hw)) {
+		ret_val = hw->mac.ops.acquire_swfw_sync(hw,
+						IXGBE_GSSR_MAC_CSR_SM);
+		if (ret_val)
+			goto out;
+
+		got_lock = true;
+	}
 
 	autoc_reg &= ~IXGBE_AUTOC_FLU;
 	autoc_reg |= IXGBE_AUTOC_AN_RESTART;
 	IXGBE_WRITE_REG(hw, IXGBE_AUTOC, autoc_reg);
 
+	if (hw->mac.type == ixgbe_mac_82599EB)
+		ixgbe_reset_pipeline_82599(hw);
+
+	if (got_lock)
+		hw->mac.ops.release_swfw_sync(hw, IXGBE_GSSR_MAC_CSR_SM);
+
 	led_reg &= ~IXGBE_LED_MODE_MASK(index);
 	led_reg &= ~IXGBE_LED_BLINK(index);
 	led_reg |= IXGBE_LED_LINK_ACTIVE << IXGBE_LED_MODE_SHIFT(index);
 	IXGBE_WRITE_REG(hw, IXGBE_LEDCTL, led_reg);
 	IXGBE_WRITE_FLUSH(hw);
 
-	return 0;
+out:
+	return ret_val;
 }
 
 /**
-- 
1.7.11.7

[net-next 13/13] ixgbe: Fix possible memory leak in ixgbe_set_ringparam

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

We were not correctly freeing the temporary rings on error in
ixgbe_set_ring_param.  In order to correct this I am unwinding a number of
changes that were made in order to get things back to the original working
form with modification for the current ring layouts.

This approach has multiple advantages including a smaller memory footprint,
and the fact that the interface is stopped while we are allocating the rings
meaning that there is less potential for some sort of memory corruption on the
ring.

The only disadvantage I see with this approach is that on a Rx allocation
failure we will report an error and only update the Tx rings.  However the
adapter should be fully functional in this state and the likelihood of such
an error is very low.  In addition it is not unreasonable to expect the
user to need to recheck the ring configuration should they experience an
error setting the ring sizes.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Phil Schmitt <phillip.j.schmitt@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c | 102 +++++++++++------------
 1 file changed, 50 insertions(+), 52 deletions(-)

diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
index 56b20d1..872c337 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ethtool.c
@@ -887,24 +887,23 @@ static int ixgbe_set_ringparam(struct net_device *netdev,
                                struct ethtool_ringparam *ring)
 {
 	struct ixgbe_adapter *adapter = netdev_priv(netdev);
-	struct ixgbe_ring *temp_tx_ring, *temp_rx_ring;
+	struct ixgbe_ring *temp_ring;
 	int i, err = 0;
 	u32 new_rx_count, new_tx_count;
-	bool need_update = false;
 
 	if ((ring->rx_mini_pending) || (ring->rx_jumbo_pending))
 		return -EINVAL;
 
-	new_rx_count = max_t(u32, ring->rx_pending, IXGBE_MIN_RXD);
-	new_rx_count = min_t(u32, new_rx_count, IXGBE_MAX_RXD);
-	new_rx_count = ALIGN(new_rx_count, IXGBE_REQ_RX_DESCRIPTOR_MULTIPLE);
-
-	new_tx_count = max_t(u32, ring->tx_pending, IXGBE_MIN_TXD);
-	new_tx_count = min_t(u32, new_tx_count, IXGBE_MAX_TXD);
+	new_tx_count = clamp_t(u32, ring->tx_pending,
+			       IXGBE_MIN_TXD, IXGBE_MAX_TXD);
 	new_tx_count = ALIGN(new_tx_count, IXGBE_REQ_TX_DESCRIPTOR_MULTIPLE);
 
-	if ((new_tx_count == adapter->tx_ring[0]->count) &&
-	    (new_rx_count == adapter->rx_ring[0]->count)) {
+	new_rx_count = clamp_t(u32, ring->rx_pending,
+			       IXGBE_MIN_RXD, IXGBE_MAX_RXD);
+	new_rx_count = ALIGN(new_rx_count, IXGBE_REQ_RX_DESCRIPTOR_MULTIPLE);
+
+	if ((new_tx_count == adapter->tx_ring_count) &&
+	    (new_rx_count == adapter->rx_ring_count)) {
 		/* nothing to do */
 		return 0;
 	}
@@ -922,81 +921,80 @@ static int ixgbe_set_ringparam(struct net_device *netdev,
 		goto clear_reset;
 	}
 
-	temp_tx_ring = vmalloc(adapter->num_tx_queues * sizeof(struct ixgbe_ring));
-	if (!temp_tx_ring) {
+	/* allocate temporary buffer to store rings in */
+	i = max_t(int, adapter->num_tx_queues, adapter->num_rx_queues);
+	temp_ring = vmalloc(i * sizeof(struct ixgbe_ring));
+
+	if (!temp_ring) {
 		err = -ENOMEM;
 		goto clear_reset;
 	}
 
+	ixgbe_down(adapter);
+
+	/*
+	 * Setup new Tx resources and free the old Tx resources in that order.
+	 * We can then assign the new resources to the rings via a memcpy.
+	 * The advantage to this approach is that we are guaranteed to still
+	 * have resources even in the case of an allocation failure.
+	 */
 	if (new_tx_count != adapter->tx_ring_count) {
 		for (i = 0; i < adapter->num_tx_queues; i++) {
-			memcpy(&temp_tx_ring[i], adapter->tx_ring[i],
+			memcpy(&temp_ring[i], adapter->tx_ring[i],
 			       sizeof(struct ixgbe_ring));
-			temp_tx_ring[i].count = new_tx_count;
-			err = ixgbe_setup_tx_resources(&temp_tx_ring[i]);
+
+			temp_ring[i].count = new_tx_count;
+			err = ixgbe_setup_tx_resources(&temp_ring[i]);
 			if (err) {
 				while (i) {
 					i--;
-					ixgbe_free_tx_resources(&temp_tx_ring[i]);
+					ixgbe_free_tx_resources(&temp_ring[i]);
 				}
-				goto clear_reset;
+				goto err_setup;
 			}
 		}
-		need_update = true;
-	}
 
-	temp_rx_ring = vmalloc(adapter->num_rx_queues * sizeof(struct ixgbe_ring));
-	if (!temp_rx_ring) {
-		err = -ENOMEM;
-		goto err_setup;
+		for (i = 0; i < adapter->num_tx_queues; i++) {
+			ixgbe_free_tx_resources(adapter->tx_ring[i]);
+
+			memcpy(adapter->tx_ring[i], &temp_ring[i],
+			       sizeof(struct ixgbe_ring));
+		}
+
+		adapter->tx_ring_count = new_tx_count;
 	}
 
+	/* Repeat the process for the Rx rings if needed */
 	if (new_rx_count != adapter->rx_ring_count) {
 		for (i = 0; i < adapter->num_rx_queues; i++) {
-			memcpy(&temp_rx_ring[i], adapter->rx_ring[i],
+			memcpy(&temp_ring[i], adapter->rx_ring[i],
 			       sizeof(struct ixgbe_ring));
-			temp_rx_ring[i].count = new_rx_count;
-			err = ixgbe_setup_rx_resources(&temp_rx_ring[i]);
+
+			temp_ring[i].count = new_rx_count;
+			err = ixgbe_setup_rx_resources(&temp_ring[i]);
 			if (err) {
 				while (i) {
 					i--;
-					ixgbe_free_rx_resources(&temp_rx_ring[i]);
+					ixgbe_free_rx_resources(&temp_ring[i]);
 				}
 				goto err_setup;
 			}
+
 		}
-		need_update = true;
-	}
 
-	/* if rings need to be updated, here's the place to do it in one shot */
-	if (need_update) {
-		ixgbe_down(adapter);
+		for (i = 0; i < adapter->num_rx_queues; i++) {
+			ixgbe_free_rx_resources(adapter->rx_ring[i]);
 
-		/* tx */
-		if (new_tx_count != adapter->tx_ring_count) {
-			for (i = 0; i < adapter->num_tx_queues; i++) {
-				ixgbe_free_tx_resources(adapter->tx_ring[i]);
-				memcpy(adapter->tx_ring[i], &temp_tx_ring[i],
-				       sizeof(struct ixgbe_ring));
-			}
-			adapter->tx_ring_count = new_tx_count;
+			memcpy(adapter->rx_ring[i], &temp_ring[i],
+			       sizeof(struct ixgbe_ring));
 		}
 
-		/* rx */
-		if (new_rx_count != adapter->rx_ring_count) {
-			for (i = 0; i < adapter->num_rx_queues; i++) {
-				ixgbe_free_rx_resources(adapter->rx_ring[i]);
-				memcpy(adapter->rx_ring[i], &temp_rx_ring[i],
-				       sizeof(struct ixgbe_ring));
-			}
-			adapter->rx_ring_count = new_rx_count;
-		}
-		ixgbe_up(adapter);
+		adapter->rx_ring_count = new_rx_count;
 	}
 
-	vfree(temp_rx_ring);
 err_setup:
-	vfree(temp_tx_ring);
+	ixgbe_up(adapter);
+	vfree(temp_ring);
 clear_reset:
 	clear_bit(__IXGBE_RESETTING, &adapter->state);
 	return err;
-- 
1.7.11.7

Re: [net-next 01/13] ixgbe: Add support for IPv6 and UDP to ixgbe_get_headlen

[Eric Dumazet]

On Fri, 2012-10-19 at 23:25 -0700, Jeff Kirsher wrote:
> From: Alexander Duyck <alexander.h.duyck@intel.com>
> 
> This change adds support for IPv6 and UDP to ixgbe_get_headlen. The
> advantage to this is that we can now handle ipv4/UDP, ipv6/TCP, and
> ipv6/UDP with a single memcpy instead of having to do them in multiple
> pskb_may_pull calls.
> 
> A quick bit of testing shows that we increase throughput for a single
> session of netperf from 8800Mpbs to about 9300Mpbs in the case of ipv6/TCP.
> As such overall ipv6 performance should improve with this change.
> 
> Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
> Tested-by: Stephen Ko  <stephen.s.ko@intel.com>
> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
> ---
>  drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 15 ++++++++++++++-
>  1 file changed, 14 insertions(+), 1 deletion(-)

Thats pretty interesting.

Why this work is not done generically, so that all drivers can have
benefit of it ?

Even a driver not using fragments could have benefit, because we could
prefetch() so that GRO stack is faster (no stall to get the last 2 bytes
of a TCP header when timestamps are used)

Thanks

Re: [net-next 01/13] ixgbe: Add support for IPv6 and UDP to ixgbe_get_headlen

[Alexander Duyck]

On 10/20/2012 1:01 AM, Eric Dumazet wrote:
> On Fri, 2012-10-19 at 23:25 -0700, Jeff Kirsher wrote:
>> From: Alexander Duyck <alexander.h.duyck@intel.com>
>>
>> This change adds support for IPv6 and UDP to ixgbe_get_headlen. The
>> advantage to this is that we can now handle ipv4/UDP, ipv6/TCP, and
>> ipv6/UDP with a single memcpy instead of having to do them in multiple
>> pskb_may_pull calls.
>>
>> A quick bit of testing shows that we increase throughput for a single
>> session of netperf from 8800Mpbs to about 9300Mpbs in the case of ipv6/TCP.
>> As such overall ipv6 performance should improve with this change.
>>
>> Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
>> Tested-by: Stephen Ko  <stephen.s.ko@intel.com>
>> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
>> ---
>>   drivers/net/ethernet/intel/ixgbe/ixgbe_main.c | 15 ++++++++++++++-
>>   1 file changed, 14 insertions(+), 1 deletion(-)
> Thats pretty interesting.
>
> Why this work is not done generically, so that all drivers can have
> benefit of it ?
>
> Even a driver not using fragments could have benefit, because we could
> prefetch() so that GRO stack is faster (no stall to get the last 2 bytes
> of a TCP header when timestamps are used)
>
> Thanks

The ixgbe_get_headlen function originally was written to handle the case 
where we were doing the hardware LRO/RSC feature but we weren't setting 
gso_size.  We were using it to get the size of the header so we could 
determine the average payload size of the frame.  When we switched over 
to page based receive the function came along with it and had actually 
prevented a performance regression for IPv4/TCP which this patch is 
fixing for the other supported protocols.  The main idea of this code is 
now to emulate what our hardware was doing in determining the length of 
the header when we had header-data split enabled.  I hadn't really 
thought of any way it could be used in a generic sense.  The function is 
essentially is built around several assumptions that only apply to the 
igb/ixgbe case such as the fact that the entire block being linear and 
the limitation of the protocols so that they match up with what we used 
to do for header-data split.

That being said, if there is a way for you to use it in the generic case 
feel free to copy the ixgbe version, rename it, and export it for use.  
The key bit for me is that I just need to be able to get the header 
length before calling eth_type_trans as I only want to do one memcpy for 
the entire header.

Thanks,

Alex

Re: [net-next 12/13] ixgbe: Add support for pipeline reset

[Martin Josefsson]

On Sat, Oct 20, 2012 at 8:26 AM, Jeff Kirsher
<jeffrey.t.kirsher@intel.com> wrote:
> From: Don Skidmore <donald.c.skidmore@intel.com>

Hi Jeff, Don,

> Calling the ixgbe_reset_pipeline_82599 function will ensure a full pipeline
> reset on all 82599 devices.  This is necessary to avoid possible link issues.
> Since this patch accomplishes this by modifying AUTOC.LMS we need to wrap
> all AUTOC writes when LESM is enabled.

The section below doesn't look right.
Old behaviour gave LMS == 011b (SFI)
New behaviour gives LMS == 000b (1 GbE link (no backplane auto-negotiation))

LMS is reset to 0 before ixgbe_reset_pipeline_82599() is called.
Looks like that part of the old code wasn't removed.

> @@ -137,28 +135,37 @@ static s32 ixgbe_setup_sfp_modules_82599(struct ixgbe_hw *hw)
>                 usleep_range(hw->eeprom.semaphore_delay * 1000,
>                              hw->eeprom.semaphore_delay * 2000);
>
> +               /* Need SW/FW semaphore around AUTOC writes if LESM on,
> +                * likewise reset_pipeline requires lock as it also writes
> +                * AUTOC.
> +                */
> +               if (ixgbe_verify_lesm_fw_enabled_82599(hw)) {
> +                       ret_val = hw->mac.ops.acquire_swfw_sync(hw,
> +                                                       IXGBE_GSSR_MAC_CSR_SM);
> +                       if (ret_val)
> +                               goto setup_sfp_out;
> +
> +                       got_lock = true;
> +               }
> +
>                 /* Now restart DSP by setting Restart_AN and clearing LMS */
>                 IXGBE_WRITE_REG(hw, IXGBE_AUTOC, ((IXGBE_READ_REG(hw,
>                                 IXGBE_AUTOC) & ~IXGBE_AUTOC_LMS_MASK) |
>                                 IXGBE_AUTOC_AN_RESTART));
>
> -               /* Wait for AN to leave state 0 */
> -               for (i = 0; i < 10; i++) {
> -                       usleep_range(4000, 8000);
> -                       reg_anlp1 = IXGBE_READ_REG(hw, IXGBE_ANLP1);
> -                       if (reg_anlp1 & IXGBE_ANLP1_AN_STATE_MASK)
> -                               break;
> +               ret_val = ixgbe_reset_pipeline_82599(hw);
> +
> +               if (got_lock) {
> +                       hw->mac.ops.release_swfw_sync(hw,
> +                                                     IXGBE_GSSR_MAC_CSR_SM);
> +                       got_lock = false;
>                 }
> -               if (!(reg_anlp1 & IXGBE_ANLP1_AN_STATE_MASK)) {
> -                       hw_dbg(hw, "sfp module setup not complete\n");
> +
> +               if (ret_val) {
> +                       hw_dbg(hw, " sfp module setup not complete\n");
>                         ret_val = IXGBE_ERR_SFP_SETUP_NOT_COMPLETE;
>                         goto setup_sfp_out;
>                 }
> -
> -               /* Restart DSP by setting Restart_AN and return to SFI mode */
> -               IXGBE_WRITE_REG(hw, IXGBE_AUTOC, (IXGBE_READ_REG(hw,
> -                               IXGBE_AUTOC) | IXGBE_AUTOC_LMS_10G_SERIAL |
> -                               IXGBE_AUTOC_AN_RESTART));
>         }
>
>  setup_sfp_out:

-- 
/Martin

RE: [net-next 12/13] ixgbe: Add support for pipeline reset

[Skidmore, Donald C]

-----Original Message-----
From: netdev-owner@vger.kernel.org [mailto:netdev-owner@vger.kernel.org] On Behalf Of Martin Josefsson
Sent: Sunday, October 21, 2012 5:06 AM
To: Kirsher, Jeffrey T
Cc: davem@davemloft.net; Skidmore, Donald C; netdev@vger.kernel.org; gospo@redhat.com; sassmann@redhat.com
Subject: Re: [net-next 12/13] ixgbe: Add support for pipeline reset

On Sat, Oct 20, 2012 at 8:26 AM, Jeff Kirsher <jeffrey.t.kirsher@intel.com> wrote:
> From: Don Skidmore <donald.c.skidmore@intel.com>

Hi Jeff, Don,

> Calling the ixgbe_reset_pipeline_82599 function will ensure a full 
> pipeline reset on all 82599 devices.  This is necessary to avoid possible link issues.
> Since this patch accomplishes this by modifying AUTOC.LMS we need to 
> wrap all AUTOC writes when LESM is enabled.

The section below doesn't look right.
Old behaviour gave LMS == 011b (SFI)
New behaviour gives LMS == 000b (1 GbE link (no backplane auto-negotiation))

LMS is reset to 0 before ixgbe_reset_pipeline_82599() is called.
Looks like that part of the old code wasn't removed.

> @@ -137,28 +135,37 @@ static s32 ixgbe_setup_sfp_modules_82599(struct ixgbe_hw *hw)
>                 usleep_range(hw->eeprom.semaphore_delay * 1000,
>                              hw->eeprom.semaphore_delay * 2000);
>
> +               /* Need SW/FW semaphore around AUTOC writes if LESM on,
> +                * likewise reset_pipeline requires lock as it also writes
> +                * AUTOC.
> +                */
> +               if (ixgbe_verify_lesm_fw_enabled_82599(hw)) {
> +                       ret_val = hw->mac.ops.acquire_swfw_sync(hw,
> +                                                       IXGBE_GSSR_MAC_CSR_SM);
> +                       if (ret_val)
> +                               goto setup_sfp_out;
> +
> +                       got_lock = true;
> +               }
> +
>                 /* Now restart DSP by setting Restart_AN and clearing LMS */
>                 IXGBE_WRITE_REG(hw, IXGBE_AUTOC, ((IXGBE_READ_REG(hw,
>                                 IXGBE_AUTOC) & ~IXGBE_AUTOC_LMS_MASK) |
>                                 IXGBE_AUTOC_AN_RESTART));
>
> -               /* Wait for AN to leave state 0 */
> -               for (i = 0; i < 10; i++) {
> -                       usleep_range(4000, 8000);
> -                       reg_anlp1 = IXGBE_READ_REG(hw, IXGBE_ANLP1);
> -                       if (reg_anlp1 & IXGBE_ANLP1_AN_STATE_MASK)
> -                               break;
> +               ret_val = ixgbe_reset_pipeline_82599(hw);
> +
> +               if (got_lock) {
> +                       hw->mac.ops.release_swfw_sync(hw,
> +                                                     IXGBE_GSSR_MAC_CSR_SM);
> +                       got_lock = false;
>                 }
> -               if (!(reg_anlp1 & IXGBE_ANLP1_AN_STATE_MASK)) {
> -                       hw_dbg(hw, "sfp module setup not complete\n");
> +
> +               if (ret_val) {
> +                       hw_dbg(hw, " sfp module setup not 
> + complete\n");
>                         ret_val = IXGBE_ERR_SFP_SETUP_NOT_COMPLETE;
>                         goto setup_sfp_out;
>                 }
> -
> -               /* Restart DSP by setting Restart_AN and return to SFI mode */
> -               IXGBE_WRITE_REG(hw, IXGBE_AUTOC, (IXGBE_READ_REG(hw,
> -                               IXGBE_AUTOC) | IXGBE_AUTOC_LMS_10G_SERIAL |
> -                               IXGBE_AUTOC_AN_RESTART));
>         }
>
>  setup_sfp_out:

--
/Martin
--
To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at  http://vger.kernel.org/majordomo-info.html

Hi Martin,

Thanks for pointing this problem out and I believe I see what is wrong.  I'll work with Jeff to get a new patch created and tested.

Thanks again,
-Don Skidmore <donald.c.skidmore@intel.com>