From: Ivan Boule <ivan.bo...@6wind.com>
The support of jumbo frames in the ixgbevf Poll Mode Driver of 10GbE
82599 VF functions consists in the following enhancements:
- Implement the mtu_set function in the ixgbevf PMD, using the IXGBE_VF_SET_LPE
request of the version 1.0 of the VF/PF mailbox API for this purpose.
- Implement the mtu_get function in the ixgbevf PMD for the sake of coherency.
- Add a detailed explanation on the VF/PF rx max frame len negotiation.
- To deal with Jumbo frames, force the receive function to handle scattered
packets.
Signed-off-by: Ivan Boule <ivan.boule at 6wind.com>
Signed-off-by: David Marchand <david.marchand at 6wind.com>
---
lib/librte_pmd_ixgbe/ixgbe_ethdev.c | 45 +++++++++++++++++++++++++++++++++++
lib/librte_pmd_ixgbe/ixgbe_rxtx.c | 18 +++++++++++++-
2 files changed, 62 insertions(+), 1 deletion(-)
diff --git a/lib/librte_pmd_ixgbe/ixgbe_ethdev.c
b/lib/librte_pmd_ixgbe/ixgbe_ethdev.c
index b9db1f4..230b758 100644
--- a/lib/librte_pmd_ixgbe/ixgbe_ethdev.c
+++ b/lib/librte_pmd_ixgbe/ixgbe_ethdev.c
@@ -194,6 +194,9 @@ static void ixgbevf_add_mac_addr(struct rte_eth_dev *dev,
uint32_t index, uint32_t pool);
static void ixgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
+static int ixgbevf_dev_get_mtu(struct rte_eth_dev *dev, uint16_t *mtu);
+static int ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t *mtu);
+
/*
* Define VF Stats MACRO for Non "cleared on read" register
*/
@@ -334,6 +337,8 @@ static struct eth_dev_ops ixgbevf_eth_dev_ops = {
.stats_reset = ixgbevf_dev_stats_reset,
.dev_close = ixgbevf_dev_close,
.dev_infos_get = ixgbe_dev_info_get,
+ .mtu_get = ixgbevf_dev_get_mtu,
+ .mtu_set = ixgbevf_dev_set_mtu,
.vlan_filter_set = ixgbevf_vlan_filter_set,
.vlan_strip_queue_set = ixgbevf_vlan_strip_queue_set,
.vlan_offload_set = ixgbevf_vlan_offload_set,
@@ -3319,6 +3324,46 @@ ixgbevf_remove_mac_addr(struct rte_eth_dev *dev,
uint32_t index)
}
}
+static int
+ixgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t *mtu)
+{
+ struct ixgbe_hw *hw;
+ uint16_t max_frame = *mtu + ETHER_HDR_LEN + ETHER_CRC_LEN;
+
+ hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+
+ /* MTU < 68 is an error and causes problems on some kernels */
+ if ((*mtu < 68) || (max_frame > ETHER_MAX_JUMBO_FRAME_LEN))
+ return -EINVAL;
+
+ /*
+ * When supported by the underlying PF driver, use the IXGBE_VF_SET_MTU
+ * request of the version 2.0 of the mailbox API.
+ * For now, use the IXGBE_VF_SET_LPE request of the version 1.0
+ * of the mailbox API.
+ * This call to IXGBE_SET_LPE action won't work with ixgbe pf drivers
+ * prior to 3.11.33 which contains the following change:
+ * "ixgbe: Enable jumbo frames support w/ SR-IOV"
+ */
+ ixgbevf_rlpml_set_vf(hw, max_frame);
+
+ /* update max frame size */
+ dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame;
+ return 0;
+}
+
+static int
+ixgbevf_dev_get_mtu(struct rte_eth_dev *dev, uint16_t *mtu)
+{
+ /*
+ * When available, use the IXGBE_VF_GET_MTU request
+ * of the version 2.0 of the mailbox API.
+ */
+ *mtu = (uint16_t) (dev->data->dev_conf.rxmode.max_rx_pkt_len -
+ (ETHER_HDR_LEN + ETHER_CRC_LEN));
+ return 0;
+}
+
static struct rte_driver rte_ixgbe_driver = {
.type = PMD_PDEV,
.init = rte_ixgbe_pmd_init,
diff --git a/lib/librte_pmd_ixgbe/ixgbe_rxtx.c
b/lib/librte_pmd_ixgbe/ixgbe_rxtx.c
index b05c1ba..c797616 100644
--- a/lib/librte_pmd_ixgbe/ixgbe_rxtx.c
+++ b/lib/librte_pmd_ixgbe/ixgbe_rxtx.c
@@ -3700,7 +3700,20 @@ ixgbevf_dev_rx_init(struct rte_eth_dev *dev)
PMD_INIT_FUNC_TRACE();
hw = IXGBE_DEV_PRIVATE_TO_HW(dev->data->dev_private);
- /* setup MTU */
+ /*
+ * When the VF driver issues a IXGBE_VF_RESET request, the PF driver
+ * disables the VF receipt of packets if the PF MTU is > 1500.
+ * This is done to deal with 82599 limitations that imposes
+ * the PF and all VFs to share the same MTU.
+ * Then, the PF driver enables again the VF receipt of packet when
+ * the VF driver issues a IXGBE_VF_SET_LPE request.
+ * In the meantime, the VF device cannot be used, even if the VF driver
+ * and the Guest VM network stack are ready to accept packets with a
+ * size up to the PF MTU.
+ * As a work-around to this PF behaviour, force the call to
+ * ixgbevf_rlpml_set_vf even if jumbo frames are not used. This way,
+ * VF packets received can work in all cases.
+ */
ixgbevf_rlpml_set_vf(hw,
(uint16_t)dev->data->dev_conf.rxmode.max_rx_pkt_len);
@@ -3783,6 +3796,9 @@ ixgbevf_dev_rx_init(struct rte_eth_dev *dev)
}
}
+ /* Set RX function to deal with Jumbo frames */
+ ixgbe_dev_set_rx_scatter_mode(dev);
+
return 0;
}
--
1.7.10.4
