net: thunderx: Support for upto 96 queues for a VF
This patch adds support for handling multiple qsets assigned to a
single VF. There by increasing no of queues from earlier 8 to max
no of CPUs in the system i.e 48 queues on a single node and 96 on
dual node system. User doesn't have option to assign which Qsets/VFs
to be merged. Upon request from VF, PF assigns next free Qsets as
secondary qsets. To maintain current behavior no of queues is kept
to 8 by default which can be increased via ethtool.
If user wants to unbind NICVF driver from a secondary Qset then it
should be done after tearing down primary VF's interface.
Signed-off-by: Sunil Goutham <sgoutham@cavium.com>
Signed-off-by: Aleksey Makarov <aleksey.makarov@caviumnetworks.com>
Signed-off-by: Robert Richter <rrichter@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/drivers/net/ethernet/cavium/thunder/nicvf_main.c b/drivers/net/ethernet/cavium/thunder/nicvf_main.c
index 85213f3..68b2dce 100644
--- a/drivers/net/ethernet/cavium/thunder/nicvf_main.c
+++ b/drivers/net/ethernet/cavium/thunder/nicvf_main.c
@@ -51,6 +51,14 @@
MODULE_PARM_DESC(cpi_alg,
"PFC algorithm (0=none, 1=VLAN, 2=VLAN16, 3=IP Diffserv)");
+static inline u8 nicvf_netdev_qidx(struct nicvf *nic, u8 qidx)
+{
+ if (nic->sqs_mode)
+ return qidx + ((nic->sqs_id + 1) * MAX_CMP_QUEUES_PER_QS);
+ else
+ return qidx;
+}
+
static inline void nicvf_set_rx_frame_cnt(struct nicvf *nic,
struct sk_buff *skb)
{
@@ -193,6 +201,7 @@
if (!nic->set_mac_pending)
ether_addr_copy(nic->netdev->dev_addr,
mbx.nic_cfg.mac_addr);
+ nic->sqs_mode = mbx.nic_cfg.sqs_mode;
nic->link_up = false;
nic->duplex = 0;
nic->speed = 0;
@@ -230,6 +239,26 @@
netif_tx_stop_all_queues(nic->netdev);
}
break;
+ case NIC_MBOX_MSG_ALLOC_SQS:
+ nic->sqs_count = mbx.sqs_alloc.qs_count;
+ nic->pf_acked = true;
+ break;
+ case NIC_MBOX_MSG_SNICVF_PTR:
+ /* Primary VF: make note of secondary VF's pointer
+ * to be used while packet transmission.
+ */
+ nic->snicvf[mbx.nicvf.sqs_id] =
+ (struct nicvf *)mbx.nicvf.nicvf;
+ nic->pf_acked = true;
+ break;
+ case NIC_MBOX_MSG_PNICVF_PTR:
+ /* Secondary VF/Qset: make note of primary VF's pointer
+ * to be used while packet reception, to handover packet
+ * to primary VF's netdev.
+ */
+ nic->pnicvf = (struct nicvf *)mbx.nicvf.nicvf;
+ nic->pf_acked = true;
+ break;
default:
netdev_err(nic->netdev,
"Invalid message from PF, msg 0x%x\n", mbx.msg.msg);
@@ -338,11 +367,100 @@
for (idx = 0; idx < rss->rss_size; idx++)
rss->ind_tbl[idx] = ethtool_rxfh_indir_default(idx,
- nic->qs->rq_cnt);
+ nic->rx_queues);
nicvf_config_rss(nic);
return 1;
}
+/* Request PF to allocate additional Qsets */
+static void nicvf_request_sqs(struct nicvf *nic)
+{
+ union nic_mbx mbx = {};
+ int sqs;
+ int sqs_count = nic->sqs_count;
+ int rx_queues = 0, tx_queues = 0;
+
+ /* Only primary VF should request */
+ if (nic->sqs_mode || !nic->sqs_count)
+ return;
+
+ mbx.sqs_alloc.msg = NIC_MBOX_MSG_ALLOC_SQS;
+ mbx.sqs_alloc.vf_id = nic->vf_id;
+ mbx.sqs_alloc.qs_count = nic->sqs_count;
+ if (nicvf_send_msg_to_pf(nic, &mbx)) {
+ /* No response from PF */
+ nic->sqs_count = 0;
+ return;
+ }
+
+ /* Return if no Secondary Qsets available */
+ if (!nic->sqs_count)
+ return;
+
+ if (nic->rx_queues > MAX_RCV_QUEUES_PER_QS)
+ rx_queues = nic->rx_queues - MAX_RCV_QUEUES_PER_QS;
+ if (nic->tx_queues > MAX_SND_QUEUES_PER_QS)
+ tx_queues = nic->tx_queues - MAX_SND_QUEUES_PER_QS;
+
+ /* Set no of Rx/Tx queues in each of the SQsets */
+ for (sqs = 0; sqs < nic->sqs_count; sqs++) {
+ mbx.nicvf.msg = NIC_MBOX_MSG_SNICVF_PTR;
+ mbx.nicvf.vf_id = nic->vf_id;
+ mbx.nicvf.sqs_id = sqs;
+ nicvf_send_msg_to_pf(nic, &mbx);
+
+ nic->snicvf[sqs]->sqs_id = sqs;
+ if (rx_queues > MAX_RCV_QUEUES_PER_QS) {
+ nic->snicvf[sqs]->qs->rq_cnt = MAX_RCV_QUEUES_PER_QS;
+ rx_queues -= MAX_RCV_QUEUES_PER_QS;
+ } else {
+ nic->snicvf[sqs]->qs->rq_cnt = rx_queues;
+ rx_queues = 0;
+ }
+
+ if (tx_queues > MAX_SND_QUEUES_PER_QS) {
+ nic->snicvf[sqs]->qs->sq_cnt = MAX_SND_QUEUES_PER_QS;
+ tx_queues -= MAX_SND_QUEUES_PER_QS;
+ } else {
+ nic->snicvf[sqs]->qs->sq_cnt = tx_queues;
+ tx_queues = 0;
+ }
+
+ nic->snicvf[sqs]->qs->cq_cnt =
+ max(nic->snicvf[sqs]->qs->rq_cnt, nic->snicvf[sqs]->qs->sq_cnt);
+
+ /* Initialize secondary Qset's queues and its interrupts */
+ nicvf_open(nic->snicvf[sqs]->netdev);
+ }
+
+ /* Update stack with actual Rx/Tx queue count allocated */
+ if (sqs_count != nic->sqs_count)
+ nicvf_set_real_num_queues(nic->netdev,
+ nic->tx_queues, nic->rx_queues);
+}
+
+/* Send this Qset's nicvf pointer to PF.
+ * PF inturn sends primary VF's nicvf struct to secondary Qsets/VFs
+ * so that packets received by these Qsets can use primary VF's netdev
+ */
+static void nicvf_send_vf_struct(struct nicvf *nic)
+{
+ union nic_mbx mbx = {};
+
+ mbx.nicvf.msg = NIC_MBOX_MSG_NICVF_PTR;
+ mbx.nicvf.sqs_mode = nic->sqs_mode;
+ mbx.nicvf.nicvf = (u64)nic;
+ nicvf_send_msg_to_pf(nic, &mbx);
+}
+
+static void nicvf_get_primary_vf_struct(struct nicvf *nic)
+{
+ union nic_mbx mbx = {};
+
+ mbx.nicvf.msg = NIC_MBOX_MSG_PNICVF_PTR;
+ nicvf_send_msg_to_pf(nic, &mbx);
+}
+
int nicvf_set_real_num_queues(struct net_device *netdev,
int tx_queues, int rx_queues)
{
@@ -453,6 +571,15 @@
struct sk_buff *skb;
struct nicvf *nic = netdev_priv(netdev);
int err = 0;
+ int rq_idx;
+
+ rq_idx = nicvf_netdev_qidx(nic, cqe_rx->rq_idx);
+
+ if (nic->sqs_mode) {
+ /* Use primary VF's 'nicvf' struct */
+ nic = nic->pnicvf;
+ netdev = nic->netdev;
+ }
/* Check for errors */
err = nicvf_check_cqe_rx_errs(nic, cq, cqe_rx);
@@ -482,7 +609,7 @@
nicvf_set_rxhash(netdev, cqe_rx, skb);
- skb_record_rx_queue(skb, cqe_rx->rq_idx);
+ skb_record_rx_queue(skb, rq_idx);
if (netdev->hw_features & NETIF_F_RXCSUM) {
/* HW by default verifies TCP/UDP/SCTP checksums */
skb->ip_summed = CHECKSUM_UNNECESSARY;
@@ -578,8 +705,11 @@
done:
/* Wakeup TXQ if its stopped earlier due to SQ full */
if (tx_done) {
- txq = netdev_get_tx_queue(netdev, cq_idx);
- if (netif_tx_queue_stopped(txq)) {
+ netdev = nic->pnicvf->netdev;
+ txq = netdev_get_tx_queue(netdev,
+ nicvf_netdev_qidx(nic, cq_idx));
+ nic = nic->pnicvf;
+ if (netif_tx_queue_stopped(txq) && netif_carrier_ok(netdev)) {
netif_tx_start_queue(txq);
nic->drv_stats.txq_wake++;
if (netif_msg_tx_err(nic))
@@ -893,7 +1023,6 @@
netdev_warn(netdev,
"%s: Transmit ring full, stopping SQ%d\n",
netdev->name, qid);
-
return NETDEV_TX_BUSY;
}
@@ -926,6 +1055,17 @@
nicvf_send_msg_to_pf(nic, &mbx);
netif_carrier_off(netdev);
+ netif_tx_stop_all_queues(nic->netdev);
+
+ /* Teardown secondary qsets first */
+ if (!nic->sqs_mode) {
+ for (qidx = 0; qidx < nic->sqs_count; qidx++) {
+ if (!nic->snicvf[qidx])
+ continue;
+ nicvf_stop(nic->snicvf[qidx]->netdev);
+ nic->snicvf[qidx] = NULL;
+ }
+ }
/* Disable RBDR & QS error interrupts */
for (qidx = 0; qidx < qs->rbdr_cnt; qidx++) {
@@ -973,6 +1113,10 @@
nicvf_free_cq_poll(nic);
+ /* Clear multiqset info */
+ nic->pnicvf = nic;
+ nic->sqs_count = 0;
+
return 0;
}
@@ -1028,10 +1172,16 @@
/* Configure CPI alorithm */
nic->cpi_alg = cpi_alg;
- nicvf_config_cpi(nic);
+ if (!nic->sqs_mode)
+ nicvf_config_cpi(nic);
+
+ nicvf_request_sqs(nic);
+ if (nic->sqs_mode)
+ nicvf_get_primary_vf_struct(nic);
/* Configure receive side scaling */
- nicvf_rss_init(nic);
+ if (!nic->sqs_mode)
+ nicvf_rss_init(nic);
err = nicvf_register_interrupts(nic);
if (err)
@@ -1282,8 +1432,7 @@
struct device *dev = &pdev->dev;
struct net_device *netdev;
struct nicvf *nic;
- struct queue_set *qs;
- int err;
+ int err, qcount;
err = pci_enable_device(pdev);
if (err) {
@@ -1309,9 +1458,17 @@
goto err_release_regions;
}
- netdev = alloc_etherdev_mqs(sizeof(struct nicvf),
- MAX_RCV_QUEUES_PER_QS,
- MAX_SND_QUEUES_PER_QS);
+ qcount = MAX_CMP_QUEUES_PER_QS;
+
+ /* Restrict multiqset support only for host bound VFs */
+ if (pdev->is_virtfn) {
+ /* Set max number of queues per VF */
+ qcount = roundup(num_online_cpus(), MAX_CMP_QUEUES_PER_QS);
+ qcount = min(qcount,
+ (MAX_SQS_PER_VF + 1) * MAX_CMP_QUEUES_PER_QS);
+ }
+
+ netdev = alloc_etherdev_mqs(sizeof(struct nicvf), qcount, qcount);
if (!netdev) {
err = -ENOMEM;
goto err_release_regions;
@@ -1324,6 +1481,8 @@
nic = netdev_priv(netdev);
nic->netdev = netdev;
nic->pdev = pdev;
+ nic->pnicvf = nic;
+ nic->max_queues = qcount;
/* MAP VF's configuration registers */
nic->reg_base = pcim_iomap(pdev, PCI_CFG_REG_BAR_NUM, 0);
@@ -1337,20 +1496,26 @@
if (err)
goto err_free_netdev;
- qs = nic->qs;
-
- err = nicvf_set_real_num_queues(netdev, qs->sq_cnt, qs->rq_cnt);
- if (err)
- goto err_free_netdev;
-
/* Check if PF is alive and get MAC address for this VF */
err = nicvf_register_misc_interrupt(nic);
if (err)
goto err_free_netdev;
+ nicvf_send_vf_struct(nic);
+
+ /* Check if this VF is in QS only mode */
+ if (nic->sqs_mode)
+ return 0;
+
+ err = nicvf_set_real_num_queues(netdev, nic->tx_queues, nic->rx_queues);
+ if (err)
+ goto err_unregister_interrupts;
+
netdev->hw_features = (NETIF_F_RXCSUM | NETIF_F_IP_CSUM | NETIF_F_SG |
NETIF_F_TSO | NETIF_F_GRO |
- NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_RXHASH);
+ NETIF_F_HW_VLAN_CTAG_RX);
+
+ netdev->hw_features |= NETIF_F_RXHASH;
netdev->features |= netdev->hw_features;
@@ -1389,8 +1554,13 @@
{
struct net_device *netdev = pci_get_drvdata(pdev);
struct nicvf *nic = netdev_priv(netdev);
+ struct net_device *pnetdev = nic->pnicvf->netdev;
- unregister_netdev(netdev);
+ /* Check if this Qset is assigned to different VF.
+ * If yes, clean primary and all secondary Qsets.
+ */
+ if (pnetdev && (pnetdev->reg_state == NETREG_REGISTERED))
+ unregister_netdev(pnetdev);
nicvf_unregister_interrupts(nic);
pci_set_drvdata(pdev, NULL);
free_netdev(netdev);