Merge branch 'bnx2x-next'

Yuval Mintz says:

====================
bnx2x: Enhancement patch series

This patch series introduces the ability to propagate link parameters
to VFs as well as control the VF link via hypervisor.

In addition, it contains 2 small improvements [one IOV-related and the
other improves performance on machines with short cache lines].

Please consider applying these patches to `net-next'.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/broadcom/bnx2x/bnx2x.h

@@ -1482,6 +1482,7 @@ struct bnx2x {
 	union pf_vf_bulletin   *pf2vf_bulletin;
 	dma_addr_t		pf2vf_bulletin_mapping;
 
+	union pf_vf_bulletin		shadow_bulletin;
 	struct pf_vf_bulletin_content	old_bulletin;
 
 	u16 requested_nr_virtfn;
@@ -1507,8 +1508,10 @@ struct bnx2x {
 /* TCP with Timestamp Option (32) + IPv6 (40) */
 #define ETH_MAX_TPA_HEADER_SIZE		72
 
-	/* Max supported alignment is 256 (8 shift) */
-#define BNX2X_RX_ALIGN_SHIFT		min(8, L1_CACHE_SHIFT)
+	/* Max supported alignment is 256 (8 shift)
+	 * minimal alignment shift 6 is optimal for 57xxx HW performance
+	 */
+#define BNX2X_RX_ALIGN_SHIFT		max(6, min(8, L1_CACHE_SHIFT))
 
 	/* FW uses 2 Cache lines Alignment for start packet and size
 	 *
@@ -1928,6 +1931,8 @@ struct bnx2x {
 	struct semaphore			stats_sema;
 
 	u8					phys_port_id[ETH_ALEN];
+
+	struct bnx2x_link_report_data		vf_link_vars;
 };
 
 /* Tx queues may be less or equal to Rx queues */

drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c

@@ -1185,29 +1185,38 @@ u16 bnx2x_get_mf_speed(struct bnx2x *bp)
 static void bnx2x_fill_report_data(struct bnx2x *bp,
 				   struct bnx2x_link_report_data *data)
 {
-	u16 line_speed = bnx2x_get_mf_speed(bp);
-
 	memset(data, 0, sizeof(*data));
 
-	/* Fill the report data: effective line speed */
-	data->line_speed = line_speed;
-
-	/* Link is down */
-	if (!bp->link_vars.link_up || (bp->flags & MF_FUNC_DIS))
-		__set_bit(BNX2X_LINK_REPORT_LINK_DOWN,
-			  &data->link_report_flags);
-
-	/* Full DUPLEX */
-	if (bp->link_vars.duplex == DUPLEX_FULL)
-		__set_bit(BNX2X_LINK_REPORT_FD, &data->link_report_flags);
-
-	/* Rx Flow Control is ON */
-	if (bp->link_vars.flow_ctrl & BNX2X_FLOW_CTRL_RX)
-		__set_bit(BNX2X_LINK_REPORT_RX_FC_ON, &data->link_report_flags);
-
-	/* Tx Flow Control is ON */
-	if (bp->link_vars.flow_ctrl & BNX2X_FLOW_CTRL_TX)
-		__set_bit(BNX2X_LINK_REPORT_TX_FC_ON, &data->link_report_flags);
+	if (IS_PF(bp)) {
+		/* Fill the report data: effective line speed */
+		data->line_speed = bnx2x_get_mf_speed(bp);
+
+		/* Link is down */
+		if (!bp->link_vars.link_up || (bp->flags & MF_FUNC_DIS))
+			__set_bit(BNX2X_LINK_REPORT_LINK_DOWN,
+				  &data->link_report_flags);
+
+		if (!BNX2X_NUM_ETH_QUEUES(bp))
+			__set_bit(BNX2X_LINK_REPORT_LINK_DOWN,
+				  &data->link_report_flags);
+
+		/* Full DUPLEX */
+		if (bp->link_vars.duplex == DUPLEX_FULL)
+			__set_bit(BNX2X_LINK_REPORT_FD,
+				  &data->link_report_flags);
+
+		/* Rx Flow Control is ON */
+		if (bp->link_vars.flow_ctrl & BNX2X_FLOW_CTRL_RX)
+			__set_bit(BNX2X_LINK_REPORT_RX_FC_ON,
+				  &data->link_report_flags);
+
+		/* Tx Flow Control is ON */
+		if (bp->link_vars.flow_ctrl & BNX2X_FLOW_CTRL_TX)
+			__set_bit(BNX2X_LINK_REPORT_TX_FC_ON,
+				  &data->link_report_flags);
+	} else { /* VF */
+		*data = bp->vf_link_vars;
+	}
 }
 
 /**
@@ -1261,6 +1270,10 @@ void __bnx2x_link_report(struct bnx2x *bp)
 	 */
 	memcpy(&bp->last_reported_link, &cur_data, sizeof(cur_data));
 
+	/* propagate status to VFs */
+	if (IS_PF(bp))
+		bnx2x_iov_link_update(bp);
+
 	if (test_bit(BNX2X_LINK_REPORT_LINK_DOWN,
 		     &cur_data.link_report_flags)) {
 		netif_carrier_off(bp->dev);

drivers/net/ethernet/broadcom/bnx2x/bnx2x_ethtool.c

@@ -216,6 +216,43 @@ static int bnx2x_get_port_type(struct bnx2x *bp)
 	return port_type;
 }
 
+static int bnx2x_get_vf_settings(struct net_device *dev,
+				 struct ethtool_cmd *cmd)
+{
+	struct bnx2x *bp = netdev_priv(dev);
+
+	if (bp->state == BNX2X_STATE_OPEN) {
+		if (test_bit(BNX2X_LINK_REPORT_FD,
+			     &bp->vf_link_vars.link_report_flags))
+			cmd->duplex = DUPLEX_FULL;
+		else
+			cmd->duplex = DUPLEX_HALF;
+
+		ethtool_cmd_speed_set(cmd, bp->vf_link_vars.line_speed);
+	} else {
+		cmd->duplex = DUPLEX_UNKNOWN;
+		ethtool_cmd_speed_set(cmd, SPEED_UNKNOWN);
+	}
+
+	cmd->port		= PORT_OTHER;
+	cmd->phy_address	= 0;
+	cmd->transceiver	= XCVR_INTERNAL;
+	cmd->autoneg		= AUTONEG_DISABLE;
+	cmd->maxtxpkt		= 0;
+	cmd->maxrxpkt		= 0;
+
+	DP(BNX2X_MSG_ETHTOOL, "ethtool_cmd: cmd %d\n"
+	   "  supported 0x%x  advertising 0x%x  speed %u\n"
+	   "  duplex %d  port %d  phy_address %d  transceiver %d\n"
+	   "  autoneg %d  maxtxpkt %d  maxrxpkt %d\n",
+	   cmd->cmd, cmd->supported, cmd->advertising,
+	   ethtool_cmd_speed(cmd),
+	   cmd->duplex, cmd->port, cmd->phy_address, cmd->transceiver,
+	   cmd->autoneg, cmd->maxtxpkt, cmd->maxrxpkt);
+
+	return 0;
+}
+
 static int bnx2x_get_settings(struct net_device *dev, struct ethtool_cmd *cmd)
 {
 	struct bnx2x *bp = netdev_priv(dev);
@@ -1110,6 +1147,10 @@ static u32 bnx2x_get_link(struct net_device *dev)
 	if (bp->flags & MF_FUNC_DIS || (bp->state != BNX2X_STATE_OPEN))
 		return 0;
 
+	if (IS_VF(bp))
+		return !test_bit(BNX2X_LINK_REPORT_LINK_DOWN,
+				 &bp->vf_link_vars.link_report_flags);
+
 	return bp->link_vars.link_up;
 }
 
@@ -3484,8 +3525,7 @@ static const struct ethtool_ops bnx2x_ethtool_ops = {
 };
 
 static const struct ethtool_ops bnx2x_vf_ethtool_ops = {
-	.get_settings		= bnx2x_get_settings,
-	.set_settings		= bnx2x_set_settings,
+	.get_settings		= bnx2x_get_vf_settings,
 	.get_drvinfo		= bnx2x_get_drvinfo,
 	.get_msglevel		= bnx2x_get_msglevel,
 	.set_msglevel		= bnx2x_set_msglevel,

drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c

@@ -2698,6 +2698,14 @@ void bnx2x__link_status_update(struct bnx2x *bp)
 		bp->link_vars.duplex = DUPLEX_FULL;
 		bp->link_vars.flow_ctrl = BNX2X_FLOW_CTRL_NONE;
 		__bnx2x_link_report(bp);
+
+		bnx2x_sample_bulletin(bp);
+
+		/* if bulletin board did not have an update for link status
+		 * __bnx2x_link_report will report current status
+		 * but it will NOT duplicate report in case of already reported
+		 * during sampling bulletin board.
+		 */
 		bnx2x_stats_handle(bp, STATS_EVENT_LINK_UP);
 	}
 }
@@ -12424,6 +12432,7 @@ static const struct net_device_ops bnx2x_netdev_ops = {
 	.ndo_busy_poll		= bnx2x_low_latency_recv,
 #endif
 	.ndo_get_phys_port_id	= bnx2x_get_phys_port_id,
+	.ndo_set_vf_link_state	= bnx2x_set_vf_link_state,
 };
 
 static int bnx2x_set_coherency_mask(struct bnx2x *bp)

drivers/net/ethernet/broadcom/bnx2x/bnx2x_sriov.h

@@ -126,7 +126,11 @@ struct bnx2x_virtf {
 #define VF_CACHE_LINE		0x0010
 #define VF_CFG_VLAN		0x0020
 #define VF_CFG_STATS_COALESCE	0x0040
-
+#define VF_CFG_EXT_BULLETIN	0x0080
+	u8 link_cfg;		/* IFLA_VF_LINK_STATE_AUTO
+				 * IFLA_VF_LINK_STATE_ENABLE
+				 * IFLA_VF_LINK_STATE_DISABLE
+				 */
 	u8 state;
 #define VF_FREE		0	/* VF ready to be acquired holds no resc */
 #define VF_ACQUIRED	1	/* VF acquired, but not initialized */
@@ -295,22 +299,22 @@ struct bnx2x_vfdb {
 #define BP_VFDB(bp)		((bp)->vfdb)
 	/* vf array */
 	struct bnx2x_virtf	*vfs;
-#define BP_VF(bp, idx)		(&((bp)->vfdb->vfs[(idx)]))
-#define bnx2x_vf(bp, idx, var)	((bp)->vfdb->vfs[(idx)].var)
+#define BP_VF(bp, idx)		(&((bp)->vfdb->vfs[idx]))
+#define bnx2x_vf(bp, idx, var)	((bp)->vfdb->vfs[idx].var)
 
 	/* queue array - for all vfs */
 	struct bnx2x_vf_queue *vfqs;
 
 	/* vf HW contexts */
 	struct hw_dma		context[BNX2X_VF_CIDS/ILT_PAGE_CIDS];
-#define	BP_VF_CXT_PAGE(bp, i)	(&(bp)->vfdb->context[(i)])
+#define	BP_VF_CXT_PAGE(bp, i)	(&(bp)->vfdb->context[i])
 
 	/* SR-IOV information */
 	struct bnx2x_sriov	sriov;
 	struct hw_dma		mbx_dma;
 #define BP_VF_MBX_DMA(bp)	(&((bp)->vfdb->mbx_dma))
 	struct bnx2x_vf_mbx	mbxs[BNX2X_MAX_NUM_OF_VFS];
-#define BP_VF_MBX(bp, vfid)	(&((bp)->vfdb->mbxs[(vfid)]))
+#define BP_VF_MBX(bp, vfid)	(&((bp)->vfdb->mbxs[vfid]))
 
 	struct hw_dma		bulletin_dma;
 #define BP_VF_BULLETIN_DMA(bp)	(&((bp)->vfdb->bulletin_dma))
@@ -336,6 +340,9 @@ struct bnx2x_vfdb {
 	/* sp_rtnl synchronization */
 	struct mutex			event_mutex;
 	u64				event_occur;
+
+	/* bulletin board update synchronization */
+	struct mutex			bulletin_mutex;
 };
 
 /* queue access */
@@ -467,9 +474,10 @@ void bnx2x_vf_handle_flr_event(struct bnx2x *bp);
 
 bool bnx2x_tlv_supported(u16 tlvtype);
 
-u32 bnx2x_crc_vf_bulletin(struct bnx2x *bp,
-			  struct pf_vf_bulletin_content *bulletin);
+u32 bnx2x_crc_vf_bulletin(struct pf_vf_bulletin_content *bulletin);
 int bnx2x_post_vf_bulletin(struct bnx2x *bp, int vf);
+void bnx2x_vf_bulletin_finalize(struct pf_vf_bulletin_content *bulletin,
+				bool support_long);
 
 enum sample_bulletin_result bnx2x_sample_bulletin(struct bnx2x *bp);
 
@@ -520,6 +528,11 @@ void bnx2x_iov_task(struct work_struct *work);
 
 void bnx2x_schedule_iov_task(struct bnx2x *bp, enum bnx2x_iov_flag flag);
 
+void bnx2x_iov_link_update(struct bnx2x *bp);
+int bnx2x_iov_link_update_vf(struct bnx2x *bp, int idx);
+
+int bnx2x_set_vf_link_state(struct net_device *dev, int vf, int link_state);
+
 #else /* CONFIG_BNX2X_SRIOV */
 
 static inline void bnx2x_iov_set_queue_sp_obj(struct bnx2x *bp, int vf_cid,
@@ -579,6 +592,14 @@ static inline void bnx2x_iov_channel_down(struct bnx2x *bp) {}
 
 static inline void bnx2x_iov_task(struct work_struct *work) {}
 static inline void bnx2x_schedule_iov_task(struct bnx2x *bp, enum bnx2x_iov_flag flag) {}
+static inline void bnx2x_iov_link_update(struct bnx2x *bp) {}
+static inline int bnx2x_iov_link_update_vf(struct bnx2x *bp, int idx) {return 0; }
+
+static inline int bnx2x_set_vf_link_state(struct net_device *dev, int vf,
+					  int link_state) {return 0; }
+struct pf_vf_bulletin_content;
+static inline void bnx2x_vf_bulletin_finalize(struct pf_vf_bulletin_content *bulletin,
+					      bool support_long) {}
 
 #endif /* CONFIG_BNX2X_SRIOV */
 #endif /* bnx2x_sriov.h */

drivers/net/ethernet/broadcom/bnx2x/bnx2x_vfpf.c

@@ -251,6 +251,9 @@ int bnx2x_vfpf_acquire(struct bnx2x *bp, u8 tx_count, u8 rx_count)
 	bnx2x_add_tlv(bp, req, req->first_tlv.tl.length,
 		      CHANNEL_TLV_PHYS_PORT_ID, sizeof(struct channel_tlv));
 
+	/* Bulletin support for bulletin board with length > legacy length */
+	req->vfdev_info.caps |= VF_CAP_SUPPORT_EXT_BULLETIN;
+
 	/* add list termination tlv */
 	bnx2x_add_tlv(bp, req,
 		      req->first_tlv.tl.length + sizeof(struct channel_tlv),
@@ -1232,6 +1235,41 @@ static void bnx2x_vf_mbx_acquire_resp(struct bnx2x *bp, struct bnx2x_virtf *vf,
 	bnx2x_vf_mbx_resp_send_msg(bp, vf, vfop_status);
 }
 
+static bool bnx2x_vf_mbx_is_windows_vm(struct bnx2x *bp,
+				       struct vfpf_acquire_tlv *acquire)
+{
+	/* Windows driver does one of three things:
+	 * 1. Old driver doesn't have bulletin board address set.
+	 * 2. 'Middle' driver sends mc_num == 32.
+	 * 3. New driver sets the OS field.
+	 */
+	if (!acquire->bulletin_addr ||
+	    acquire->resc_request.num_mc_filters == 32 ||
+	    ((acquire->vfdev_info.vf_os & VF_OS_MASK) ==
+	     VF_OS_WINDOWS))
+		return true;
+
+	return false;
+}
+
+static int bnx2x_vf_mbx_acquire_chk_dorq(struct bnx2x *bp,
+					 struct bnx2x_virtf *vf,
+					 struct bnx2x_vf_mbx *mbx)
+{
+	/* Linux drivers which correctly set the doorbell size also
+	 * send a physical port request
+	 */
+	if (bnx2x_search_tlv_list(bp, &mbx->msg->req,
+				  CHANNEL_TLV_PHYS_PORT_ID))
+		return 0;
+
+	/* Issue does not exist in windows VMs */
+	if (bnx2x_vf_mbx_is_windows_vm(bp, &mbx->msg->req.acquire))
+		return 0;
+
+	return -EOPNOTSUPP;
+}
+
 static void bnx2x_vf_mbx_acquire(struct bnx2x *bp, struct bnx2x_virtf *vf,
 				 struct bnx2x_vf_mbx *mbx)
 {
@@ -1247,12 +1285,32 @@ static void bnx2x_vf_mbx_acquire(struct bnx2x *bp, struct bnx2x_virtf *vf,
 	   acquire->resc_request.num_vlan_filters,
 	   acquire->resc_request.num_mc_filters);
 
+	/* Prevent VFs with old drivers from loading, since they calculate
+	 * CIDs incorrectly requiring a VF-flr [VM reboot] in order to recover
+	 * while being upgraded.
+	 */
+	rc = bnx2x_vf_mbx_acquire_chk_dorq(bp, vf, mbx);
+	if (rc) {
+		DP(BNX2X_MSG_IOV,
+		   "VF [%d] - Can't support acquire request due to doorbell mismatch. Please update VM driver\n",
+		   vf->abs_vfid);
+		goto out;
+	}
+
 	/* acquire the resources */
 	rc = bnx2x_vf_acquire(bp, vf, &acquire->resc_request);
 
 	/* store address of vf's bulletin board */
 	vf->bulletin_map = acquire->bulletin_addr;
+	if (acquire->vfdev_info.caps & VF_CAP_SUPPORT_EXT_BULLETIN) {
+		DP(BNX2X_MSG_IOV, "VF[%d] supports long bulletin boards\n",
+		   vf->abs_vfid);
+		vf->cfg_flags |= VF_CFG_EXT_BULLETIN;
+	} else {
+		vf->cfg_flags &= ~VF_CFG_EXT_BULLETIN;
+	}
 
+out:
 	/* response */
 	bnx2x_vf_mbx_acquire_resp(bp, vf, mbx, rc);
 }
@@ -1273,6 +1331,10 @@ static void bnx2x_vf_mbx_init_vf(struct bnx2x *bp, struct bnx2x_virtf *vf,
 	if (init->flags & VFPF_INIT_FLG_STATS_COALESCE)
 		vf->cfg_flags |= VF_CFG_STATS_COALESCE;
 
+	/* Update VF's view of link state */
+	if (vf->cfg_flags & VF_CFG_EXT_BULLETIN)
+		bnx2x_iov_link_update_vf(bp, vf->index);
+
 	/* response */
 	bnx2x_vf_mbx_resp(bp, vf, rc);
 }
@@ -2007,6 +2069,17 @@ void bnx2x_vf_mbx(struct bnx2x *bp)
 	}
 }
 
+void bnx2x_vf_bulletin_finalize(struct pf_vf_bulletin_content *bulletin,
+				bool support_long)
+{
+	/* Older VFs contain a bug where they can't check CRC for bulletin
+	 * boards of length greater than legacy size.
+	 */
+	bulletin->length = support_long ? BULLETIN_CONTENT_SIZE :
+					  BULLETIN_CONTENT_LEGACY_SIZE;
+	bulletin->crc = bnx2x_crc_vf_bulletin(bulletin);
+}
+
 /* propagate local bulletin board to vf */
 int bnx2x_post_vf_bulletin(struct bnx2x *bp, int vf)
 {
@@ -2023,8 +2096,9 @@ int bnx2x_post_vf_bulletin(struct bnx2x *bp, int vf)
 
 	/* increment bulletin board version and compute crc */
 	bulletin->version++;
-	bulletin->length = BULLETIN_CONTENT_SIZE;
-	bulletin->crc = bnx2x_crc_vf_bulletin(bp, bulletin);
+	bnx2x_vf_bulletin_finalize(bulletin,
+				   (bnx2x_vf(bp, vf, cfg_flags) &
+				    VF_CFG_EXT_BULLETIN) ? true : false);
 
 	/* propagate bulletin board via dmae to vm memory */
 	rc = bnx2x_copy32_vf_dmae(bp, false, pf_addr,

drivers/net/ethernet/broadcom/bnx2x/bnx2x_vfpf.h

@@ -65,6 +65,7 @@ struct hw_sb_info {
 #define VFPF_RX_MASK_ACCEPT_ALL_MULTICAST	0x00000008
 #define VFPF_RX_MASK_ACCEPT_BROADCAST		0x00000010
 #define BULLETIN_CONTENT_SIZE		(sizeof(struct pf_vf_bulletin_content))
+#define BULLETIN_CONTENT_LEGACY_SIZE	(32)
 #define BULLETIN_ATTEMPTS	5 /* crc failures before throwing towel */
 #define BULLETIN_CRC_SEED	0
 
@@ -117,7 +118,15 @@ struct vfpf_acquire_tlv {
 		/* the following fields are for debug purposes */
 		u8  vf_id;		/* ME register value */
 		u8  vf_os;		/* e.g. Linux, W2K8 */
-		u8 padding[2];
+#define VF_OS_SUBVERSION_MASK	(0x1f)
+#define VF_OS_MASK		(0xe0)
+#define VF_OS_SHIFT		(5)
+#define VF_OS_UNDEFINED		(0 << VF_OS_SHIFT)
+#define VF_OS_WINDOWS		(1 << VF_OS_SHIFT)
+
+		u8 padding;
+		u8 caps;
+#define VF_CAP_SUPPORT_EXT_BULLETIN	(1 << 0)
 	} vfdev_info;
 
 	struct vf_pf_resc_request resc_request;
@@ -393,11 +402,23 @@ struct pf_vf_bulletin_content {
 					 * to attempt to send messages on the
 					 * channel after this bit is set
 					 */
+#define LINK_VALID		3	/* alert the VF thet a new link status
+					 * update is available for it
+					 */
 	u8 mac[ETH_ALEN];
 	u8 mac_padding[2];
 
 	u16 vlan;
 	u8 vlan_padding[6];
+
+	u16 link_speed;			 /* Effective line speed */
+	u8 link_speed_padding[6];
+	u32 link_flags;			 /* VFPF_LINK_REPORT_XXX flags */
+#define VFPF_LINK_REPORT_LINK_DOWN	 (1 << 0)
+#define VFPF_LINK_REPORT_FULL_DUPLEX	 (1 << 1)
+#define VFPF_LINK_REPORT_RX_FC_ON	 (1 << 2)
+#define VFPF_LINK_REPORT_TX_FC_ON	 (1 << 3)
+	u8 link_flags_padding[4];
 };
 
 union pf_vf_bulletin {