ixgbe: Add SR-IOV specific features
authorGreg Rose <gregory.v.rose@intel.com>
Sat, 9 Jan 2010 02:25:48 +0000 (02:25 +0000)
committerDavid S. Miller <davem@davemloft.net>
Sun, 10 Jan 2010 21:34:25 +0000 (13:34 -0800)
This module and header file add functions to support SR-IOV features of the
82599 10Gbe controller.

Signed-off-by: Greg Rose <gregory.v.rose@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ixgbe/ixgbe_sriov.c [new file with mode: 0644]
drivers/net/ixgbe/ixgbe_sriov.h [new file with mode: 0644]

diff --git a/drivers/net/ixgbe/ixgbe_sriov.c b/drivers/net/ixgbe/ixgbe_sriov.c
new file mode 100644 (file)
index 0000000..74bca74
--- /dev/null
@@ -0,0 +1,336 @@
+/*******************************************************************************
+
+  Intel 10 Gigabit PCI Express Linux driver
+  Copyright(c) 1999 - 2009 Intel Corporation.
+
+  This program is free software; you can redistribute it and/or modify it
+  under the terms and conditions of the GNU General Public License,
+  version 2, as published by the Free Software Foundation.
+
+  This program is distributed in the hope it will be useful, but WITHOUT
+  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+  more details.
+
+  You should have received a copy of the GNU General Public License along with
+  this program; if not, write to the Free Software Foundation, Inc.,
+  51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+
+  The full GNU General Public License is included in this distribution in
+  the file called "COPYING".
+
+  Contact Information:
+  e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
+  Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
+
+*******************************************************************************/
+
+
+#include <linux/types.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/netdevice.h>
+#include <linux/vmalloc.h>
+#include <linux/string.h>
+#include <linux/in.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/ipv6.h>
+#ifdef NETIF_F_HW_VLAN_TX
+#include <linux/if_vlan.h>
+#endif
+
+#include "ixgbe.h"
+
+#include "ixgbe_sriov.h"
+
+int ixgbe_set_vf_multicasts(struct ixgbe_adapter *adapter,
+                           int entries, u16 *hash_list, u32 vf)
+{
+       struct vf_data_storage *vfinfo = &adapter->vfinfo[vf];
+       int i;
+
+       /* only so many hash values supported */
+       entries = min(entries, IXGBE_MAX_VF_MC_ENTRIES);
+
+       /*
+        * salt away the number of multi cast addresses assigned
+        * to this VF for later use to restore when the PF multi cast
+        * list changes
+        */
+       vfinfo->num_vf_mc_hashes = entries;
+
+       /*
+        * VFs are limited to using the MTA hash table for their multicast
+        * addresses
+        */
+       for (i = 0; i < entries; i++) {
+               vfinfo->vf_mc_hashes[i] = hash_list[i];;
+       }
+
+       /* Flush and reset the mta with the new values */
+       ixgbe_set_rx_mode(adapter->netdev);
+
+       return 0;
+}
+
+void ixgbe_restore_vf_multicasts(struct ixgbe_adapter *adapter)
+{
+       struct ixgbe_hw *hw = &adapter->hw;
+       struct vf_data_storage *vfinfo;
+       int i, j;
+       u32 vector_bit;
+       u32 vector_reg;
+       u32 mta_reg;
+
+       for (i = 0; i < adapter->num_vfs; i++) {
+               vfinfo = &adapter->vfinfo[i];
+               for (j = 0; j < vfinfo->num_vf_mc_hashes; j++) {
+                       hw->addr_ctrl.mta_in_use++;
+                       vector_reg = (vfinfo->vf_mc_hashes[j] >> 5) & 0x7F;
+                       vector_bit = vfinfo->vf_mc_hashes[j] & 0x1F;
+                       mta_reg = IXGBE_READ_REG(hw, IXGBE_MTA(vector_reg));
+                       mta_reg |= (1 << vector_bit);
+                       IXGBE_WRITE_REG(hw, IXGBE_MTA(vector_reg), mta_reg);
+               }
+       }
+}
+
+int ixgbe_set_vf_vlan(struct ixgbe_adapter *adapter, int add, int vid, u32 vf)
+{
+       u32 ctrl;
+
+       /* Check if global VLAN already set, if not set it */
+       ctrl = IXGBE_READ_REG(&adapter->hw, IXGBE_VLNCTRL);
+       if (!(ctrl & IXGBE_VLNCTRL_VFE)) {
+               /* enable VLAN tag insert/strip */
+               ctrl |= IXGBE_VLNCTRL_VFE;
+               ctrl &= ~IXGBE_VLNCTRL_CFIEN;
+               IXGBE_WRITE_REG(&adapter->hw, IXGBE_VLNCTRL, ctrl);
+       }
+
+       return adapter->hw.mac.ops.set_vfta(&adapter->hw, vid, vf, (bool)add);
+}
+
+
+void ixgbe_set_vmolr(struct ixgbe_hw *hw, u32 vf)
+{
+       u32 vmolr = IXGBE_READ_REG(hw, IXGBE_VMOLR(vf));
+       vmolr |= (IXGBE_VMOLR_AUPE |
+                 IXGBE_VMOLR_ROMPE |
+                 IXGBE_VMOLR_ROPE |
+                 IXGBE_VMOLR_BAM);
+       IXGBE_WRITE_REG(hw, IXGBE_VMOLR(vf), vmolr);
+}
+
+inline void ixgbe_vf_reset_event(struct ixgbe_adapter *adapter, u32 vf)
+{
+       struct ixgbe_hw *hw = &adapter->hw;
+
+       /* reset offloads to defaults */
+       ixgbe_set_vmolr(hw, vf);
+
+
+       /* reset multicast table array for vf */
+       adapter->vfinfo[vf].num_vf_mc_hashes = 0;
+
+       /* Flush and reset the mta with the new values */
+       ixgbe_set_rx_mode(adapter->netdev);
+
+       if (adapter->vfinfo[vf].rar > 0) {
+               adapter->hw.mac.ops.clear_rar(&adapter->hw,
+                                             adapter->vfinfo[vf].rar);
+               adapter->vfinfo[vf].rar = -1;
+       }
+}
+
+int ixgbe_set_vf_mac(struct ixgbe_adapter *adapter,
+                          int vf, unsigned char *mac_addr)
+{
+       struct ixgbe_hw *hw = &adapter->hw;
+
+       adapter->vfinfo[vf].rar = hw->mac.ops.set_rar(hw, vf + 1, mac_addr,
+                                                     vf, IXGBE_RAH_AV);
+       if (adapter->vfinfo[vf].rar < 0) {
+               DPRINTK(DRV, ERR, "Could not set MAC Filter for VF %d\n", vf);
+               return -1;
+       }
+
+       memcpy(adapter->vfinfo[vf].vf_mac_addresses, mac_addr, 6);
+
+       return 0;
+}
+
+int ixgbe_vf_configuration(struct pci_dev *pdev, unsigned int event_mask)
+{
+       unsigned char vf_mac_addr[6];
+       struct net_device *netdev = pci_get_drvdata(pdev);
+       struct ixgbe_adapter *adapter = netdev_priv(netdev);
+       unsigned int vfn = (event_mask & 0x3f);
+
+       bool enable = ((event_mask & 0x10000000U) != 0);
+
+       if (enable) {
+               random_ether_addr(vf_mac_addr);
+               DPRINTK(PROBE, INFO, "IOV: VF %d is enabled "
+                      "mac %02X:%02X:%02X:%02X:%02X:%02X\n",
+                      vfn,
+                      vf_mac_addr[0], vf_mac_addr[1], vf_mac_addr[2],
+                      vf_mac_addr[3], vf_mac_addr[4], vf_mac_addr[5]);
+               /*
+                * Store away the VF "permananet" MAC address, it will ask
+                * for it later.
+                */
+               memcpy(adapter->vfinfo[vfn].vf_mac_addresses, vf_mac_addr, 6);
+       }
+
+       return 0;
+}
+
+inline void ixgbe_vf_reset_msg(struct ixgbe_adapter *adapter, u32 vf)
+{
+       struct ixgbe_hw *hw = &adapter->hw;
+       u32 reg;
+       u32 reg_offset, vf_shift;
+
+       vf_shift = vf % 32;
+       reg_offset = vf / 32;
+
+       /* enable transmit and receive for vf */
+       reg = IXGBE_READ_REG(hw, IXGBE_VFTE(reg_offset));
+       reg |= (reg | (1 << vf_shift));
+       IXGBE_WRITE_REG(hw, IXGBE_VFTE(reg_offset), reg);
+
+       reg = IXGBE_READ_REG(hw, IXGBE_VFRE(reg_offset));
+       reg |= (reg | (1 << vf_shift));
+       IXGBE_WRITE_REG(hw, IXGBE_VFRE(reg_offset), reg);
+
+       ixgbe_vf_reset_event(adapter, vf);
+}
+
+static int ixgbe_rcv_msg_from_vf(struct ixgbe_adapter *adapter, u32 vf)
+{
+       u32 mbx_size = IXGBE_VFMAILBOX_SIZE;
+       u32 msgbuf[mbx_size];
+       struct ixgbe_hw *hw = &adapter->hw;
+       s32 retval;
+       int entries;
+       u16 *hash_list;
+       int add, vid;
+
+       retval = ixgbe_read_mbx(hw, msgbuf, mbx_size, vf);
+
+       if (retval)
+               printk(KERN_ERR "Error receiving message from VF\n");
+
+       /* this is a message we already processed, do nothing */
+       if (msgbuf[0] & (IXGBE_VT_MSGTYPE_ACK | IXGBE_VT_MSGTYPE_NACK))
+               return retval;
+
+       /*
+        * until the vf completes a virtual function reset it should not be
+        * allowed to start any configuration.
+        */
+
+       if (msgbuf[0] == IXGBE_VF_RESET) {
+               unsigned char *vf_mac = adapter->vfinfo[vf].vf_mac_addresses;
+               u8 *addr = (u8 *)(&msgbuf[1]);
+               DPRINTK(PROBE, INFO, "VF Reset msg received from vf %d\n", vf);
+               adapter->vfinfo[vf].clear_to_send = false;
+               ixgbe_vf_reset_msg(adapter, vf);
+               adapter->vfinfo[vf].clear_to_send = true;
+
+               /* reply to reset with ack and vf mac address */
+               msgbuf[0] = IXGBE_VF_RESET | IXGBE_VT_MSGTYPE_ACK;
+               memcpy(addr, vf_mac, IXGBE_ETH_LENGTH_OF_ADDRESS);
+               /*
+                * Piggyback the multicast filter type so VF can compute the
+                * correct vectors
+                */
+               msgbuf[3] = hw->mac.mc_filter_type;
+               ixgbe_write_mbx(hw, msgbuf, IXGBE_VF_PERMADDR_MSG_LEN, vf);
+
+               return retval;
+       }
+
+       if (!adapter->vfinfo[vf].clear_to_send) {
+               msgbuf[0] |= IXGBE_VT_MSGTYPE_NACK;
+               ixgbe_write_mbx(hw, msgbuf, 1, vf);
+               return retval;
+       }
+
+       switch ((msgbuf[0] & 0xFFFF)) {
+       case IXGBE_VF_SET_MAC_ADDR:
+               {
+                       u8 *new_mac = ((u8 *)(&msgbuf[1]));
+                       if (is_valid_ether_addr(new_mac))
+                               ixgbe_set_vf_mac(adapter, vf, new_mac);
+                       else
+                               retval = -1;
+               }
+               break;
+       case IXGBE_VF_SET_MULTICAST:
+               entries = (msgbuf[0] & IXGBE_VT_MSGINFO_MASK)
+                         >> IXGBE_VT_MSGINFO_SHIFT;
+               hash_list = (u16 *)&msgbuf[1];
+               retval = ixgbe_set_vf_multicasts(adapter, entries,
+                                                hash_list, vf);
+               break;
+       case IXGBE_VF_SET_LPE:
+               WARN_ON((msgbuf[0] & 0xFFFF) == IXGBE_VF_SET_LPE);
+               break;
+       case IXGBE_VF_SET_VLAN:
+               add = (msgbuf[0] & IXGBE_VT_MSGINFO_MASK)
+                     >> IXGBE_VT_MSGINFO_SHIFT;
+               vid = (msgbuf[1] & IXGBE_VLVF_VLANID_MASK);
+               retval = ixgbe_set_vf_vlan(adapter, add, vid, vf);
+               break;
+       default:
+               DPRINTK(DRV, ERR, "Unhandled Msg %8.8x\n", msgbuf[0]);
+               retval = IXGBE_ERR_MBX;
+               break;
+       }
+
+       /* notify the VF of the results of what it sent us */
+       if (retval)
+               msgbuf[0] |= IXGBE_VT_MSGTYPE_NACK;
+       else
+               msgbuf[0] |= IXGBE_VT_MSGTYPE_ACK;
+
+       msgbuf[0] |= IXGBE_VT_MSGTYPE_CTS;
+
+       ixgbe_write_mbx(hw, msgbuf, 1, vf);
+
+       return retval;
+}
+
+static void ixgbe_rcv_ack_from_vf(struct ixgbe_adapter *adapter, u32 vf)
+{
+       struct ixgbe_hw *hw = &adapter->hw;
+       u32 msg = IXGBE_VT_MSGTYPE_NACK;
+
+       /* if device isn't clear to send it shouldn't be reading either */
+       if (!adapter->vfinfo[vf].clear_to_send)
+               ixgbe_write_mbx(hw, &msg, 1, vf);
+}
+
+void ixgbe_msg_task(struct ixgbe_adapter *adapter)
+{
+       struct ixgbe_hw *hw = &adapter->hw;
+       u32 vf;
+
+       for (vf = 0; vf < adapter->num_vfs; vf++) {
+               /* process any reset requests */
+               if (!ixgbe_check_for_rst(hw, vf))
+                       ixgbe_vf_reset_event(adapter, vf);
+
+               /* process any messages pending */
+               if (!ixgbe_check_for_msg(hw, vf))
+                       ixgbe_rcv_msg_from_vf(adapter, vf);
+
+               /* process any acks */
+               if (!ixgbe_check_for_ack(hw, vf))
+                       ixgbe_rcv_ack_from_vf(adapter, vf);
+       }
+}
+
diff --git a/drivers/net/ixgbe/ixgbe_sriov.h b/drivers/net/ixgbe/ixgbe_sriov.h
new file mode 100644 (file)
index 0000000..664b237
--- /dev/null
@@ -0,0 +1,45 @@
+/*******************************************************************************
+
+  Intel 10 Gigabit PCI Express Linux driver
+  Copyright(c) 1999 - 2009 Intel Corporation.
+
+  This program is free software; you can redistribute it and/or modify it
+  under the terms and conditions of the GNU General Public License,
+  version 2, as published by the Free Software Foundation.
+
+  This program is distributed in the hope it will be useful, but WITHOUT
+  ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+  FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+  more details.
+
+  You should have received a copy of the GNU General Public License along with
+  this program; if not, write to the Free Software Foundation, Inc.,
+  51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+
+  The full GNU General Public License is included in this distribution in
+  the file called "COPYING".
+
+  Contact Information:
+  e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
+  Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
+
+*******************************************************************************/
+
+#ifndef _IXGBE_SRIOV_H_
+#define _IXGBE_SRIOV_H_
+
+int ixgbe_set_vf_multicasts(struct ixgbe_adapter *adapter,
+                            int entries, u16 *hash_list, u32 vf);
+void ixgbe_restore_vf_multicasts(struct ixgbe_adapter *adapter);
+int ixgbe_set_vf_vlan(struct ixgbe_adapter *adapter, int add, int vid, u32 vf);
+void ixgbe_set_vmolr(struct ixgbe_hw *hw, u32 vf);
+void ixgbe_vf_reset_event(struct ixgbe_adapter *adapter, u32 vf);
+void ixgbe_vf_reset_msg(struct ixgbe_adapter *adapter, u32 vf);
+void ixgbe_msg_task(struct ixgbe_adapter *adapter);
+int ixgbe_set_vf_mac(struct ixgbe_adapter *adapter,
+                     int vf, unsigned char *mac_addr);
+int ixgbe_vf_configuration(struct pci_dev *pdev, unsigned int event_mask);
+void ixgbe_dump_registers(struct ixgbe_adapter *adapter);
+
+#endif /* _IXGBE_SRIOV_H_ */
+