| Message ID | 1274806509.18023.58.camel@localhost.localdomain |
|---|---|
| State | Awaiting Upstream, archived |
| Delegated to: | David Miller |
| Headers | show |
To produce this problem: 1. modprobe ixgbe max_vfs=2 eth4 is PF, eth5 is VF 2. ip link set eth4 up 3. ip link add link eth4 address 54:52:00:35:e3:20 macvlan2 type macvlan 4. ip addr add 192.168.7.74/24 dev macvlan2 5. ping macvlan2 from remote host, works 6. ip link set eth5 up 7. ping macvlan2 from remote host failed. Based on my understanding, the problem is: 1. PF set_rar use rar index is 0, and vmdq index is adapter->num_vfs, 2. when macvlan2 is created, rar index is based rar_used_count, which would be 1. 3. later when VF is up, the rar index is vf+1, and vmdq index is vf, so VF0 will overwrite macvlan2 rar entry. The fix here: 1. make sure PF uses vmdq index = adapter->num_vfs during initialization, reset. 2. reserve rar index for all VFs from 1 to num_vfs + 1. Please let me know whether my understanding is correct or not. Thanks Shirley -- 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
>-----Original Message----- >From: Shirley Ma [mailto:mashirle@us.ibm.com] >Sent: Tuesday, May 25, 2010 10:16 AM >To: Kirsher, Jeffrey T >Cc: Rose, Gregory V; davem@davemloft.net; kvm@vger.kernel.org; >netdev@vger.kernel.org; e1000-devel@lists.sourceforge.net >Subject: Re: [PATCH net-next] ixgbe: make macvlan on PF working when >SRIOV is enabled > >To produce this problem: > >1. modprobe ixgbe max_vfs=2 > eth4 is PF, eth5 is VF >2. ip link set eth4 up >3. ip link add link eth4 address 54:52:00:35:e3:20 macvlan2 type macvlan >4. ip addr add 192.168.7.74/24 dev macvlan2 >5. ping macvlan2 from remote host, works >6. ip link set eth5 up >7. ping macvlan2 from remote host failed. > >Based on my understanding, the problem is: >1. PF set_rar use rar index is 0, and vmdq index is adapter->num_vfs, >2. when macvlan2 is created, rar index is based rar_used_count, which >would be 1. >3. later when VF is up, the rar index is vf+1, and vmdq index is vf, so >VF0 will overwrite macvlan2 rar entry. > >The fix here: >1. make sure PF uses vmdq index = adapter->num_vfs during >initialization, reset. >2. reserve rar index for all VFs from 1 to num_vfs + 1. > > >Please let me know whether my understanding is correct or not. Yes, that appears to be correct. We'll test your patch but I think you're on the right track. - Greg
On Tue, May 25, 2010 at 09:55, Shirley Ma <mashirle@us.ibm.com> wrote: > When SRIOV is enabled, if we create macvlan on PF and load the VF driver in host > domain, when macvlan/PF and VF interfaces are all up, the macvlan incoming network > traffics are directed to VF interface not PF interface. This patch has fixed this > issue. > > Signed-off-by: Shirley Ma <xma@us.ibm.com> > --- > drivers/net/ixgbe/ixgbe_common.c | 26 ++++++++++++++++---------- > 1 files changed, 16 insertions(+), 10 deletions(-) > Thanks, I have added the patch to my queue.
diff --git a/drivers/net/ixgbe/ixgbe_common.c b/drivers/net/ixgbe/ixgbe_common.c index 1159d91..591dd19 100644 --- a/drivers/net/ixgbe/ixgbe_common.c +++ b/drivers/net/ixgbe/ixgbe_common.c @@ -1271,6 +1271,7 @@ s32 ixgbe_init_rx_addrs_generic(struct ixgbe_hw *hw) { u32 i; u32 rar_entries = hw->mac.num_rar_entries; + struct ixgbe_adapter *adapter = hw->back; /* * If the current mac address is valid, assume it is a software override @@ -1288,15 +1289,18 @@ s32 ixgbe_init_rx_addrs_generic(struct ixgbe_hw *hw) hw_dbg(hw, "Overriding MAC Address in RAR[0]\n"); hw_dbg(hw, " New MAC Addr =%pM\n", hw->mac.addr); - hw->mac.ops.set_rar(hw, 0, hw->mac.addr, 0, IXGBE_RAH_AV); + hw->mac.ops.set_rar(hw, 0, hw->mac.addr, adapter->num_vfs, + IXGBE_RAH_AV); } hw->addr_ctrl.overflow_promisc = 0; - hw->addr_ctrl.rar_used_count = 1; + /* reserve VFs receive address entries if any */ + hw->addr_ctrl.rar_used_count = adapter->num_vfs + 1; - /* Zero out the other receive addresses. */ - hw_dbg(hw, "Clearing RAR[1-%d]\n", rar_entries - 1); - for (i = 1; i < rar_entries; i++) { + /* Zero out the other receive addresses except VFs if any */ + hw_dbg(hw, "Clearing RAR[%d-%d]\n", + adapter->num_vfs + 1, rar_entries - 1); + for (i = adapter->num_vfs + 1; i < rar_entries; i++) { IXGBE_WRITE_REG(hw, IXGBE_RAL(i), 0); IXGBE_WRITE_REG(hw, IXGBE_RAH(i), 0); } @@ -1368,18 +1372,20 @@ s32 ixgbe_update_uc_addr_list_generic(struct ixgbe_hw *hw, u32 uc_addr_in_use; u32 fctrl; struct netdev_hw_addr *ha; + struct ixgbe_adapter *adapter = hw->back; /* * Clear accounting of old secondary address list, * don't count RAR[0] */ - uc_addr_in_use = hw->addr_ctrl.rar_used_count - 1; + uc_addr_in_use = hw->addr_ctrl.rar_used_count - 1 - adapter->num_vfs; hw->addr_ctrl.rar_used_count -= uc_addr_in_use; hw->addr_ctrl.overflow_promisc = 0; - /* Zero out the other receive addresses */ - hw_dbg(hw, "Clearing RAR[1-%d]\n", uc_addr_in_use + 1); - for (i = 0; i < uc_addr_in_use; i++) { + /* Zero out the other receive addresses except VFs if any */ + hw_dbg(hw, "Clearing RAR[%d-%d]\n", + adapter->num_vfs + 1, uc_addr_in_use + adapter->num_vfs + 1); + for (i = adapter->num_vfs; i < uc_addr_in_use + adapter->num_vfs; i++) { IXGBE_WRITE_REG(hw, IXGBE_RAL(1+i), 0); IXGBE_WRITE_REG(hw, IXGBE_RAH(1+i), 0); } @@ -1387,7 +1393,7 @@ s32 ixgbe_update_uc_addr_list_generic(struct ixgbe_hw *hw, /* Add the new addresses */ netdev_for_each_uc_addr(ha, netdev) { hw_dbg(hw, " Adding the secondary addresses:\n"); - ixgbe_add_uc_addr(hw, ha->addr, 0); + ixgbe_add_uc_addr(hw, ha->addr, adapter->num_vfs); } if (hw->addr_ctrl.overflow_promisc) {
When SRIOV is enabled, if we create macvlan on PF and load the VF driver in host domain, when macvlan/PF and VF interfaces are all up, the macvlan incoming network traffics are directed to VF interface not PF interface. This patch has fixed this issue. Signed-off-by: Shirley Ma <xma@us.ibm.com> --- drivers/net/ixgbe/ixgbe_common.c | 26 ++++++++++++++++---------- 1 files changed, 16 insertions(+), 10 deletions(-) -- 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