On Sat, 23 Dec 2023 05:39:58 -0800 Shinas Rasheed wrote:
> +static int octep_vf_napi_poll(struct napi_struct *napi, int budget)
> +{
> + struct octep_vf_ioq_vector *ioq_vector =
> + container_of(napi, struct octep_vf_ioq_vector, napi);
> + u32 tx_pending, rx_done;
> +
> + tx_pending = octep_vf_iq_process_completions(ioq_vector->iq, budget);
completions should use some static budget, say 256, the budget passed
in is for Rx.
> + rx_done = octep_vf_oq_process_rx(ioq_vector->oq, budget);
You should not call Rx processing if budget is 0 at all, please see
NAPI docs. Looks like the function may try to refill Rx buffers with
budget of 0.
> + /* need more polling if tx completion processing is still pending or
> + * processed at least 'budget' number of rx packets.
> + */
> + if (tx_pending || rx_done >= budget)
> + return budget;
> +
> + napi_complete(napi);
please use napi_complete_done().
> + octep_vf_enable_ioq_irq(ioq_vector->iq, ioq_vector->oq);
> + return rx_done;
> +}
> /**
> * octep_vf_start_xmit() - Enqueue packet to Octoen hardware Tx Queue.
> *
> @@ -184,6 +591,143 @@ static int octep_vf_stop(struct net_device *netdev)
> static netdev_tx_t octep_vf_start_xmit(struct sk_buff *skb,
> struct net_device *netdev)
> {
> [...]
> +dma_map_sg_err:
> + if (si > 0) {
> + dma_unmap_single(iq->dev, sglist[0].dma_ptr[0],
> + sglist[0].len[0], DMA_TO_DEVICE);
> + sglist[0].len[0] = 0;
> + }
> + while (si > 1) {
> + dma_unmap_page(iq->dev, sglist[si >> 2].dma_ptr[si & 3],
> + sglist[si >> 2].len[si & 3], DMA_TO_DEVICE);
> + sglist[si >> 2].len[si & 3] = 0;
> + si--;
> + }
> + tx_buffer->gather = 0;
> +dma_map_err:
if previous tx had xmit_more() you need to ring the doorbell here
> + dev_kfree_skb_any(skb);
> return NETDEV_TX_OK;
> }
> @@ -114,8 +158,8 @@ static int octep_vf_setup_oq(struct octep_vf_device *oct, int q_no)
> goto desc_dma_alloc_err;
> }
>
> - oq->buff_info = (struct octep_vf_rx_buffer *)
> - vzalloc(oq->max_count * OCTEP_VF_OQ_RECVBUF_SIZE);
> + oq->buff_info = vzalloc(oq->max_count * OCTEP_VF_OQ_RECVBUF_SIZE);
> +
bad fixup squash?
> if (unlikely(!oq->buff_info)) {
> dev_err(&oct->pdev->dev,
> "Failed to allocate buffer info for OQ-%d\n", q_no);
> + /* Swap the length field that is in Big-Endian to CPU */
> + buff_info->len = be64_to_cpu(resp_hw->length);
> + if (oct->fw_info.rx_ol_flags) {
> + /* Extended response header is immediately after
> + * response header (resp_hw)
> + */
> + resp_hw_ext = (struct octep_vf_oq_resp_hw_ext *)
> + (resp_hw + 1);
> + buff_info->len -= OCTEP_VF_OQ_RESP_HW_EXT_SIZE;
> + /* Packet Data is immediately after
> + * extended response header.
> + */
> + data_offset = OCTEP_VF_OQ_RESP_HW_SIZE +
> + OCTEP_VF_OQ_RESP_HW_EXT_SIZE;
> + rx_ol_flags = resp_hw_ext->rx_ol_flags;
> + } else {
> + /* Data is immediately after
> + * Hardware Rx response header.
> + */
> + data_offset = OCTEP_VF_OQ_RESP_HW_SIZE;
> + rx_ol_flags = 0;
> + }
> + rx_bytes += buff_info->len;
> +
> + if (buff_info->len <= oq->max_single_buffer_size) {
> + skb = build_skb((void *)resp_hw, PAGE_SIZE);
napi_build_skb() ?
> + skb_reserve(skb, data_offset);
> + skb_put(skb, buff_info->len);
> + read_idx++;
> + desc_used++;
> + if (read_idx == oq->max_count)
> + read_idx = 0;
> + } else {
> + struct skb_shared_info *shinfo;
> + u16 data_len;
> +
> + skb = build_skb((void *)resp_hw, PAGE_SIZE);
ditto
> + skb_reserve(skb, data_offset);
> + /* Head fragment includes response header(s);
> + * subsequent fragments contains only data.
> + */
> + skb_put(skb, oq->max_single_buffer_size);
> + read_idx++;
> + desc_used++;
> + if (read_idx == oq->max_count)
> + read_idx = 0;
> +
> + shinfo = skb_shinfo(skb);
> + data_len = buff_info->len - oq->max_single_buffer_size;
> + while (data_len) {
> + dma_unmap_page(oq->dev, oq->desc_ring[read_idx].buffer_ptr,
> + PAGE_SIZE, DMA_FROM_DEVICE);
> + buff_info = (struct octep_vf_rx_buffer *)
> + &oq->buff_info[read_idx];
> + if (data_len < oq->buffer_size) {
> + buff_info->len = data_len;
> + data_len = 0;
> + } else {
> + buff_info->len = oq->buffer_size;
> + data_len -= oq->buffer_size;
> + }
> +
> + skb_add_rx_frag(skb, shinfo->nr_frags,
> + buff_info->page, 0,
> + buff_info->len,
> + buff_info->len);
> + buff_info->page = NULL;
> + read_idx++;
> + desc_used++;
> + if (read_idx == oq->max_count)
> + read_idx = 0;
> + }
> + }
> +
> + skb->dev = oq->netdev;
> + skb->protocol = eth_type_trans(skb, skb->dev);
double space
> + if (feat & NETIF_F_RXCSUM &&
> + OCTEP_VF_RX_CSUM_VERIFIED(rx_ol_flags))
> + skb->ip_summed = CHECKSUM_UNNECESSARY;
> + else
> + skb->ip_summed = CHECKSUM_NONE;
> + napi_gro_receive(oq->napi, skb);
> + }
> +
> + oq->host_read_idx = read_idx;
> + oq->refill_count += desc_used;
> + oq->stats.packets += pkt;
> + oq->stats.bytes += rx_bytes;
> +
> + return pkt;
> +}
> +/**
> + * octep_vf_iq_process_completions() - Process Tx queue completions.
> + *
> + * @iq: Octeon Tx queue data structure.
> + * @budget: max number of completions to be processed in one invocation.
> + */
> +int octep_vf_iq_process_completions(struct octep_vf_iq *iq, u16 budget)
> +{
> [...]
> + netdev_tx_completed_queue(iq->netdev_q, compl_pkts, compl_bytes);
> +
> + if (unlikely(__netif_subqueue_stopped(iq->netdev, iq->q_no)) &&
> + (IQ_INSTR_SPACE(iq) >
> + OCTEP_VF_WAKE_QUEUE_THRESHOLD))
> + netif_wake_subqueue(iq->netdev, iq->q_no);
Please use helpers form net/netdev_queues.h
> @@ -195,8 +267,7 @@ static void octep_vf_free_iq(struct octep_vf_iq *iq)
>
> desc_ring_size = OCTEP_VF_IQ_DESC_SIZE * CFG_GET_IQ_NUM_DESC(oct->conf);
>
> - if (iq->buff_info)
> - vfree(iq->buff_info);
> + vfree(iq->buff_info);
This should be in a previous patch.
@@ -36,6 +36,371 @@ MODULE_AUTHOR("Veerasenareddy Burru <vburru@marvell.com>");
MODULE_DESCRIPTION(OCTEP_VF_DRV_STRING);
MODULE_LICENSE("GPL");
+/**
+ * octep_vf_alloc_ioq_vectors() - Allocate Tx/Rx Queue interrupt info.
+ *
+ * @oct: Octeon device private data structure.
+ *
+ * Allocate resources to hold per Tx/Rx queue interrupt info.
+ * This is the information passed to interrupt handler, from which napi poll
+ * is scheduled and includes quick access to private data of Tx/Rx queue
+ * corresponding to the interrupt being handled.
+ *
+ * Return: 0, on successful allocation of resources for all queue interrupts.
+ * -1, if failed to allocate any resource.
+ */
+static int octep_vf_alloc_ioq_vectors(struct octep_vf_device *oct)
+{
+ struct octep_vf_ioq_vector *ioq_vector;
+ int i;
+
+ for (i = 0; i < oct->num_oqs; i++) {
+ oct->ioq_vector[i] = vzalloc(sizeof(*oct->ioq_vector[i]));
+ if (!oct->ioq_vector[i])
+ goto free_ioq_vector;
+
+ ioq_vector = oct->ioq_vector[i];
+ ioq_vector->iq = oct->iq[i];
+ ioq_vector->oq = oct->oq[i];
+ ioq_vector->octep_vf_dev = oct;
+ }
+
+ dev_info(&oct->pdev->dev, "Allocated %d IOQ vectors\n", oct->num_oqs);
+ return 0;
+
+free_ioq_vector:
+ while (i) {
+ i--;
+ vfree(oct->ioq_vector[i]);
+ oct->ioq_vector[i] = NULL;
+ }
+ return -1;
+}
+
+/**
+ * octep_vf_free_ioq_vectors() - Free Tx/Rx Queue interrupt vector info.
+ *
+ * @oct: Octeon device private data structure.
+ */
+static void octep_vf_free_ioq_vectors(struct octep_vf_device *oct)
+{
+ int i;
+
+ for (i = 0; i < oct->num_oqs; i++) {
+ if (oct->ioq_vector[i]) {
+ vfree(oct->ioq_vector[i]);
+ oct->ioq_vector[i] = NULL;
+ }
+ }
+ netdev_info(oct->netdev, "Freed IOQ Vectors\n");
+}
+
+/**
+ * octep_vf_enable_msix_range() - enable MSI-x interrupts.
+ *
+ * @oct: Octeon device private data structure.
+ *
+ * Allocate and enable all MSI-x interrupts (queue and non-queue interrupts)
+ * for the Octeon device.
+ *
+ * Return: 0, on successfully enabling all MSI-x interrupts.
+ * -1, if failed to enable any MSI-x interrupt.
+ */
+static int octep_vf_enable_msix_range(struct octep_vf_device *oct)
+{
+ int num_msix, msix_allocated;
+ int i;
+
+ /* Generic interrupts apart from input/output queues */
+ //num_msix = oct->num_oqs + CFG_GET_NON_IOQ_MSIX(oct->conf);
+ num_msix = oct->num_oqs;
+ oct->msix_entries = kcalloc(num_msix, sizeof(struct msix_entry), GFP_KERNEL);
+ if (!oct->msix_entries)
+ goto msix_alloc_err;
+
+ for (i = 0; i < num_msix; i++)
+ oct->msix_entries[i].entry = i;
+
+ msix_allocated = pci_enable_msix_range(oct->pdev, oct->msix_entries,
+ num_msix, num_msix);
+ if (msix_allocated != num_msix) {
+ dev_err(&oct->pdev->dev,
+ "Failed to enable %d msix irqs; got only %d\n",
+ num_msix, msix_allocated);
+ goto enable_msix_err;
+ }
+ oct->num_irqs = msix_allocated;
+ dev_info(&oct->pdev->dev, "MSI-X enabled successfully\n");
+
+ return 0;
+
+enable_msix_err:
+ if (msix_allocated > 0)
+ pci_disable_msix(oct->pdev);
+ kfree(oct->msix_entries);
+ oct->msix_entries = NULL;
+msix_alloc_err:
+ return -1;
+}
+
+/**
+ * octep_vf_disable_msix() - disable MSI-x interrupts.
+ *
+ * @oct: Octeon device private data structure.
+ *
+ * Disable MSI-x on the Octeon device.
+ */
+static void octep_vf_disable_msix(struct octep_vf_device *oct)
+{
+ pci_disable_msix(oct->pdev);
+ kfree(oct->msix_entries);
+ oct->msix_entries = NULL;
+ dev_info(&oct->pdev->dev, "Disabled MSI-X\n");
+}
+
+/**
+ * octep_vf_ioq_intr_handler() - handler for all Tx/Rx queue interrupts.
+ *
+ * @irq: Interrupt number.
+ * @data: interrupt data contains pointers to Tx/Rx queue private data
+ * and correspong NAPI context.
+ *
+ * this is common handler for all non-queue (generic) interrupts.
+ */
+static irqreturn_t octep_vf_ioq_intr_handler(int irq, void *data)
+{
+ struct octep_vf_ioq_vector *ioq_vector = data;
+ struct octep_vf_device *oct = ioq_vector->octep_vf_dev;
+
+ return oct->hw_ops.ioq_intr_handler(ioq_vector);
+}
+
+/**
+ * octep_vf_request_irqs() - Register interrupt handlers.
+ *
+ * @oct: Octeon device private data structure.
+ *
+ * Register handlers for all queue and non-queue interrupts.
+ *
+ * Return: 0, on successful registration of all interrupt handlers.
+ * -1, on any error.
+ */
+static int octep_vf_request_irqs(struct octep_vf_device *oct)
+{
+ struct net_device *netdev = oct->netdev;
+ struct octep_vf_ioq_vector *ioq_vector;
+ struct msix_entry *msix_entry;
+ int ret, i;
+
+ /* Request IRQs for Tx/Rx queues */
+ for (i = 0; i < oct->num_oqs; i++) {
+ ioq_vector = oct->ioq_vector[i];
+ msix_entry = &oct->msix_entries[i];
+
+ snprintf(ioq_vector->name, sizeof(ioq_vector->name),
+ "%s-q%d", netdev->name, i);
+ ret = request_irq(msix_entry->vector,
+ octep_vf_ioq_intr_handler, 0,
+ ioq_vector->name, ioq_vector);
+ if (ret) {
+ netdev_err(netdev,
+ "request_irq failed for Q-%d; err=%d",
+ i, ret);
+ goto ioq_irq_err;
+ }
+
+ cpumask_set_cpu(i % num_online_cpus(),
+ &ioq_vector->affinity_mask);
+ irq_set_affinity_hint(msix_entry->vector,
+ &ioq_vector->affinity_mask);
+ }
+
+ return 0;
+ioq_irq_err:
+ while (i) {
+ --i;
+ free_irq(oct->msix_entries[i].vector, oct);
+ }
+ return -1;
+}
+
+/**
+ * octep_vf_free_irqs() - free all registered interrupts.
+ *
+ * @oct: Octeon device private data structure.
+ *
+ * Free all queue and non-queue interrupts of the Octeon device.
+ */
+static void octep_vf_free_irqs(struct octep_vf_device *oct)
+{
+ int i;
+
+ for (i = 0; i < oct->num_irqs; i++) {
+ irq_set_affinity_hint(oct->msix_entries[i].vector, NULL);
+ free_irq(oct->msix_entries[i].vector, oct->ioq_vector[i]);
+ }
+ netdev_info(oct->netdev, "IRQs freed\n");
+}
+
+/**
+ * octep_vf_setup_irqs() - setup interrupts for the Octeon device.
+ *
+ * @oct: Octeon device private data structure.
+ *
+ * Allocate data structures to hold per interrupt information, allocate/enable
+ * MSI-x interrupt and register interrupt handlers.
+ *
+ * Return: 0, on successful allocation and registration of all interrupts.
+ * -1, on any error.
+ */
+static int octep_vf_setup_irqs(struct octep_vf_device *oct)
+{
+ if (octep_vf_alloc_ioq_vectors(oct))
+ goto ioq_vector_err;
+
+ if (octep_vf_enable_msix_range(oct))
+ goto enable_msix_err;
+
+ if (octep_vf_request_irqs(oct))
+ goto request_irq_err;
+
+ return 0;
+
+request_irq_err:
+ octep_vf_disable_msix(oct);
+enable_msix_err:
+ octep_vf_free_ioq_vectors(oct);
+ioq_vector_err:
+ return -1;
+}
+
+/**
+ * octep_vf_clean_irqs() - free all interrupts and its resources.
+ *
+ * @oct: Octeon device private data structure.
+ */
+static void octep_vf_clean_irqs(struct octep_vf_device *oct)
+{
+ octep_vf_free_irqs(oct);
+ octep_vf_disable_msix(oct);
+ octep_vf_free_ioq_vectors(oct);
+}
+
+/**
+ * octep_vf_enable_ioq_irq() - Enable MSI-x interrupt of a Tx/Rx queue.
+ *
+ * @iq: Octeon Tx queue data structure.
+ * @oq: Octeon Rx queue data structure.
+ */
+static void octep_vf_enable_ioq_irq(struct octep_vf_iq *iq, struct octep_vf_oq *oq)
+{
+ u32 pkts_pend = oq->pkts_pending;
+
+ netdev_dbg(iq->netdev, "enabling intr for Q-%u\n", iq->q_no);
+ if (iq->pkts_processed) {
+ writel(iq->pkts_processed, iq->inst_cnt_reg);
+ iq->pkt_in_done -= iq->pkts_processed;
+ iq->pkts_processed = 0;
+ }
+ if (oq->last_pkt_count - pkts_pend) {
+ writel(oq->last_pkt_count - pkts_pend, oq->pkts_sent_reg);
+ oq->last_pkt_count = pkts_pend;
+ }
+
+ /* Flush the previous wrties before writing to RESEND bit */
+ smp_wmb();
+ writeq(1UL << OCTEP_VF_OQ_INTR_RESEND_BIT, oq->pkts_sent_reg);
+ writeq(1UL << OCTEP_VF_IQ_INTR_RESEND_BIT, iq->inst_cnt_reg);
+}
+
+/**
+ * octep_vf_napi_poll() - NAPI poll function for Tx/Rx.
+ *
+ * @napi: pointer to napi context.
+ * @budget: max number of packets to be processed in single invocation.
+ */
+static int octep_vf_napi_poll(struct napi_struct *napi, int budget)
+{
+ struct octep_vf_ioq_vector *ioq_vector =
+ container_of(napi, struct octep_vf_ioq_vector, napi);
+ u32 tx_pending, rx_done;
+
+ tx_pending = octep_vf_iq_process_completions(ioq_vector->iq, budget);
+ rx_done = octep_vf_oq_process_rx(ioq_vector->oq, budget);
+
+ /* need more polling if tx completion processing is still pending or
+ * processed at least 'budget' number of rx packets.
+ */
+ if (tx_pending || rx_done >= budget)
+ return budget;
+
+ napi_complete(napi);
+ octep_vf_enable_ioq_irq(ioq_vector->iq, ioq_vector->oq);
+ return rx_done;
+}
+
+/**
+ * octep_vf_napi_add() - Add NAPI poll for all Tx/Rx queues.
+ *
+ * @oct: Octeon device private data structure.
+ */
+static void octep_vf_napi_add(struct octep_vf_device *oct)
+{
+ int i;
+
+ for (i = 0; i < oct->num_oqs; i++) {
+ netdev_dbg(oct->netdev, "Adding NAPI on Q-%d\n", i);
+ netif_napi_add(oct->netdev, &oct->ioq_vector[i]->napi, octep_vf_napi_poll);
+ oct->oq[i]->napi = &oct->ioq_vector[i]->napi;
+ }
+}
+
+/**
+ * octep_vf_napi_delete() - delete NAPI poll callback for all Tx/Rx queues.
+ *
+ * @oct: Octeon device private data structure.
+ */
+static void octep_vf_napi_delete(struct octep_vf_device *oct)
+{
+ int i;
+
+ for (i = 0; i < oct->num_oqs; i++) {
+ netdev_dbg(oct->netdev, "Deleting NAPI on Q-%d\n", i);
+ netif_napi_del(&oct->ioq_vector[i]->napi);
+ oct->oq[i]->napi = NULL;
+ }
+}
+
+/**
+ * octep_vf_napi_enable() - enable NAPI for all Tx/Rx queues.
+ *
+ * @oct: Octeon device private data structure.
+ */
+static void octep_vf_napi_enable(struct octep_vf_device *oct)
+{
+ int i;
+
+ for (i = 0; i < oct->num_oqs; i++) {
+ netdev_dbg(oct->netdev, "Enabling NAPI on Q-%d\n", i);
+ napi_enable(&oct->ioq_vector[i]->napi);
+ }
+}
+
+/**
+ * octep_vf_napi_disable() - disable NAPI for all Tx/Rx queues.
+ *
+ * @oct: Octeon device private data structure.
+ */
+static void octep_vf_napi_disable(struct octep_vf_device *oct)
+{
+ int i;
+
+ for (i = 0; i < oct->num_oqs; i++) {
+ netdev_dbg(oct->netdev, "Disabling NAPI on Q-%d\n", i);
+ napi_disable(&oct->ioq_vector[i]->napi);
+ }
+}
+
static void octep_vf_link_up(struct net_device *netdev)
{
netif_carrier_on(netdev);
@@ -98,6 +463,8 @@ static int octep_vf_open(struct net_device *netdev)
goto setup_iq_err;
if (octep_vf_setup_oqs(oct))
goto setup_oq_err;
+ if (octep_vf_setup_irqs(oct))
+ goto setup_irq_err;
err = netif_set_real_num_tx_queues(netdev, oct->num_oqs);
if (err)
@@ -106,6 +473,9 @@ static int octep_vf_open(struct net_device *netdev)
if (err)
goto set_queues_err;
+ octep_vf_napi_add(oct);
+ octep_vf_napi_enable(oct);
+
oct->link_info.admin_up = 1;
octep_vf_set_rx_state(oct, true);
@@ -128,6 +498,10 @@ static int octep_vf_open(struct net_device *netdev)
return 0;
set_queues_err:
+ octep_vf_napi_disable(oct);
+ octep_vf_napi_delete(oct);
+ octep_vf_clean_irqs(oct);
+setup_irq_err:
octep_vf_free_oqs(oct);
setup_oq_err:
octep_vf_free_iqs(oct);
@@ -161,7 +535,10 @@ static int octep_vf_stop(struct net_device *netdev)
oct->link_info.oper_up = 0;
oct->hw_ops.disable_interrupts(oct);
+ octep_vf_napi_disable(oct);
+ octep_vf_napi_delete(oct);
+ octep_vf_clean_irqs(oct);
octep_vf_clean_iqs(oct);
oct->hw_ops.disable_io_queues(oct);
@@ -172,6 +549,36 @@ static int octep_vf_stop(struct net_device *netdev)
return 0;
}
+/**
+ * octep_vf_iq_full_check() - check if a Tx queue is full.
+ *
+ * @iq: Octeon Tx queue data structure.
+ *
+ * Return: 0, if the Tx queue is not full.
+ * 1, if the Tx queue is full.
+ */
+static int octep_vf_iq_full_check(struct octep_vf_iq *iq)
+{
+ if (likely((IQ_INSTR_SPACE(iq)) >
+ OCTEP_VF_WAKE_QUEUE_THRESHOLD))
+ return 0;
+
+ /* Stop the queue if unable to send */
+ netif_stop_subqueue(iq->netdev, iq->q_no);
+
+ /* check again and restart the queue, in case NAPI has just freed
+ * enough Tx ring entries.
+ */
+ if (unlikely(IQ_INSTR_SPACE(iq) >
+ OCTEP_VF_WAKE_QUEUE_THRESHOLD)) {
+ netif_start_subqueue(iq->netdev, iq->q_no);
+ iq->stats.restart_cnt++;
+ return 0;
+ }
+
+ return 1;
+}
+
/**
* octep_vf_start_xmit() - Enqueue packet to Octoen hardware Tx Queue.
*
@@ -184,6 +591,143 @@ static int octep_vf_stop(struct net_device *netdev)
static netdev_tx_t octep_vf_start_xmit(struct sk_buff *skb,
struct net_device *netdev)
{
+ struct octep_vf_device *oct = netdev_priv(netdev);
+ netdev_features_t feat = netdev->features;
+ struct octep_vf_tx_sglist_desc *sglist;
+ struct octep_vf_tx_buffer *tx_buffer;
+ struct octep_vf_tx_desc_hw *hw_desc;
+ struct skb_shared_info *shinfo;
+ struct octep_vf_instr_hdr *ih;
+ struct octep_vf_iq *iq;
+ skb_frag_t *frag;
+ u16 nr_frags, si;
+ int xmit_more;
+ u16 q_no, wi;
+
+ if (skb_put_padto(skb, ETH_ZLEN))
+ return NETDEV_TX_OK;
+
+ q_no = skb_get_queue_mapping(skb);
+ if (q_no >= oct->num_iqs) {
+ netdev_err(netdev, "Invalid Tx skb->queue_mapping=%d\n", q_no);
+ q_no = q_no % oct->num_iqs;
+ }
+
+ iq = oct->iq[q_no];
+ if (octep_vf_iq_full_check(iq)) {
+ iq->stats.tx_busy++;
+ return NETDEV_TX_BUSY;
+ }
+
+ shinfo = skb_shinfo(skb);
+ nr_frags = shinfo->nr_frags;
+
+ wi = iq->host_write_index;
+ hw_desc = &iq->desc_ring[wi];
+ hw_desc->ih64 = 0;
+
+ tx_buffer = iq->buff_info + wi;
+ tx_buffer->skb = skb;
+
+ ih = &hw_desc->ih;
+ ih->tlen = skb->len;
+ ih->pkind = oct->fw_info.pkind;
+ ih->fsz = oct->fw_info.fsz;
+ ih->tlen = skb->len + ih->fsz;
+
+ if (!nr_frags) {
+ tx_buffer->gather = 0;
+ tx_buffer->dma = dma_map_single(iq->dev, skb->data,
+ skb->len, DMA_TO_DEVICE);
+ if (dma_mapping_error(iq->dev, tx_buffer->dma))
+ goto dma_map_err;
+ hw_desc->dptr = tx_buffer->dma;
+ } else {
+ /* Scatter/Gather */
+ dma_addr_t dma;
+ u16 len;
+
+ sglist = tx_buffer->sglist;
+
+ ih->gsz = nr_frags + 1;
+ ih->gather = 1;
+ tx_buffer->gather = 1;
+
+ len = skb_headlen(skb);
+ dma = dma_map_single(iq->dev, skb->data, len, DMA_TO_DEVICE);
+ if (dma_mapping_error(iq->dev, dma))
+ goto dma_map_err;
+
+ memset(sglist, 0, OCTEP_VF_SGLIST_SIZE_PER_PKT);
+ sglist[0].len[3] = len;
+ sglist[0].dma_ptr[0] = dma;
+
+ si = 1; /* entry 0 is main skb, mapped above */
+ frag = &shinfo->frags[0];
+ while (nr_frags--) {
+ len = skb_frag_size(frag);
+ dma = skb_frag_dma_map(iq->dev, frag, 0,
+ len, DMA_TO_DEVICE);
+ if (dma_mapping_error(iq->dev, dma))
+ goto dma_map_sg_err;
+
+ sglist[si >> 2].len[3 - (si & 3)] = len;
+ sglist[si >> 2].dma_ptr[si & 3] = dma;
+
+ frag++;
+ si++;
+ }
+ hw_desc->dptr = tx_buffer->sglist_dma;
+ }
+ if (oct->fw_info.tx_ol_flags) {
+ if ((feat & (NETIF_F_TSO)) && (skb_is_gso(skb))) {
+ hw_desc->txm.ol_flags = OCTEP_VF_TX_OFFLOAD_CKSUM;
+ hw_desc->txm.ol_flags |= OCTEP_VF_TX_OFFLOAD_TSO;
+ hw_desc->txm.gso_size = skb_shinfo(skb)->gso_size;
+ hw_desc->txm.gso_segs = skb_shinfo(skb)->gso_segs;
+ } else if (feat & (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)) {
+ hw_desc->txm.ol_flags = OCTEP_VF_TX_OFFLOAD_CKSUM;
+ }
+ /* due to ESR txm will be swappeed by hw */
+ hw_desc->txm64[0] = (__force u64)cpu_to_be64(hw_desc->txm64[0]);
+ }
+
+ netdev_tx_sent_queue(iq->netdev_q, skb->len);
+
+ xmit_more = netdev_xmit_more();
+
+ skb_tx_timestamp(skb);
+ iq->fill_cnt++;
+ wi++;
+ iq->host_write_index = wi & iq->ring_size_mask;
+ if (xmit_more &&
+ (IQ_INSTR_PENDING(iq) <
+ (iq->max_count - OCTEP_VF_WAKE_QUEUE_THRESHOLD)) &&
+ iq->fill_cnt < iq->fill_threshold)
+ return NETDEV_TX_OK;
+
+ /* Flush the hw descriptors before writing to doorbell */
+ smp_wmb();
+ writel(iq->fill_cnt, iq->doorbell_reg);
+ iq->stats.instr_posted += iq->fill_cnt;
+ iq->fill_cnt = 0;
+ return NETDEV_TX_OK;
+
+dma_map_sg_err:
+ if (si > 0) {
+ dma_unmap_single(iq->dev, sglist[0].dma_ptr[0],
+ sglist[0].len[0], DMA_TO_DEVICE);
+ sglist[0].len[0] = 0;
+ }
+ while (si > 1) {
+ dma_unmap_page(iq->dev, sglist[si >> 2].dma_ptr[si & 3],
+ sglist[si >> 2].len[si & 3], DMA_TO_DEVICE);
+ sglist[si >> 2].len[si & 3] = 0;
+ si--;
+ }
+ tx_buffer->gather = 0;
+dma_map_err:
+ dev_kfree_skb_any(skb);
return NETDEV_TX_OK;
}
@@ -114,7 +114,7 @@ static int __octep_vf_mbox_send_cmd(struct octep_vf_device *oct,
{
struct octep_vf_mbox *mbox = oct->mbox;
u64 reg_val = 0ull;
- int count = 0;
+ int count;
if (!mbox)
return OCTEP_PFVF_MBOX_CMD_STATUS_NOT_SETUP;
@@ -68,6 +68,50 @@ static int octep_vf_oq_fill_ring_buffers(struct octep_vf_oq *oq)
return -1;
}
+/**
+ * octep_vf_oq_refill() - refill buffers for used Rx ring descriptors.
+ *
+ * @oct: Octeon device private data structure.
+ * @oq: Octeon Rx queue data structure.
+ *
+ * Return: number of descriptors successfully refilled with receive buffers.
+ */
+static int octep_vf_oq_refill(struct octep_vf_device *oct, struct octep_vf_oq *oq)
+{
+ struct octep_vf_oq_desc_hw *desc_ring = oq->desc_ring;
+ struct page *page;
+ u32 refill_idx, i;
+
+ refill_idx = oq->host_refill_idx;
+ for (i = 0; i < oq->refill_count; i++) {
+ page = dev_alloc_page();
+ if (unlikely(!page)) {
+ dev_err(oq->dev, "refill: rx buffer alloc failed\n");
+ oq->stats.alloc_failures++;
+ break;
+ }
+
+ desc_ring[refill_idx].buffer_ptr = dma_map_page(oq->dev, page, 0,
+ PAGE_SIZE, DMA_FROM_DEVICE);
+ if (dma_mapping_error(oq->dev, desc_ring[refill_idx].buffer_ptr)) {
+ dev_err(oq->dev,
+ "OQ-%d buffer refill: DMA mapping error!\n",
+ oq->q_no);
+ put_page(page);
+ oq->stats.alloc_failures++;
+ break;
+ }
+ oq->buff_info[refill_idx].page = page;
+ refill_idx++;
+ if (refill_idx == oq->max_count)
+ refill_idx = 0;
+ }
+ oq->host_refill_idx = refill_idx;
+ oq->refill_count -= i;
+
+ return i;
+}
+
/**
* octep_vf_setup_oq() - Setup a Rx queue.
*
@@ -114,8 +158,8 @@ static int octep_vf_setup_oq(struct octep_vf_device *oct, int q_no)
goto desc_dma_alloc_err;
}
- oq->buff_info = (struct octep_vf_rx_buffer *)
- vzalloc(oq->max_count * OCTEP_VF_OQ_RECVBUF_SIZE);
+ oq->buff_info = vzalloc(oq->max_count * OCTEP_VF_OQ_RECVBUF_SIZE);
+
if (unlikely(!oq->buff_info)) {
dev_err(&oct->pdev->dev,
"Failed to allocate buffer info for OQ-%d\n", q_no);
@@ -262,3 +306,206 @@ void octep_vf_free_oqs(struct octep_vf_device *oct)
"Successfully freed OQ(RxQ)-%d.\n", i);
}
}
+
+/**
+ * octep_vf_oq_check_hw_for_pkts() - Check for new Rx packets.
+ *
+ * @oct: Octeon device private data structure.
+ * @oq: Octeon Rx queue data structure.
+ *
+ * Return: packets received after previous check.
+ */
+static int octep_vf_oq_check_hw_for_pkts(struct octep_vf_device *oct,
+ struct octep_vf_oq *oq)
+{
+ u32 pkt_count, new_pkts;
+
+ pkt_count = readl(oq->pkts_sent_reg);
+ new_pkts = pkt_count - oq->last_pkt_count;
+
+ /* Clear the hardware packets counter register if the rx queue is
+ * being processed continuously with-in a single interrupt and
+ * reached half its max value.
+ * this counter is not cleared every time read, to save write cycles.
+ */
+ if (unlikely(pkt_count > 0xF0000000U)) {
+ writel(pkt_count, oq->pkts_sent_reg);
+ pkt_count = readl(oq->pkts_sent_reg);
+ new_pkts += pkt_count;
+ }
+ oq->last_pkt_count = pkt_count;
+ oq->pkts_pending += new_pkts;
+ return new_pkts;
+}
+
+/**
+ * __octep_vf_oq_process_rx() - Process hardware Rx queue and push to stack.
+ *
+ * @oct: Octeon device private data structure.
+ * @oq: Octeon Rx queue data structure.
+ * @pkts_to_process: number of packets to be processed.
+ *
+ * Process the new packets in Rx queue.
+ * Packets larger than single Rx buffer arrive in consecutive descriptors.
+ * But, count returned by the API only accounts full packets, not fragments.
+ *
+ * Return: number of packets processed and pushed to stack.
+ */
+static int __octep_vf_oq_process_rx(struct octep_vf_device *oct,
+ struct octep_vf_oq *oq, u16 pkts_to_process)
+{
+ struct octep_vf_oq_resp_hw_ext *resp_hw_ext = NULL;
+ netdev_features_t feat = oq->netdev->features;
+ struct octep_vf_rx_buffer *buff_info;
+ struct octep_vf_oq_resp_hw *resp_hw;
+ u32 pkt, rx_bytes, desc_used;
+ u16 data_offset, rx_ol_flags;
+ struct sk_buff *skb;
+ u32 read_idx;
+
+ read_idx = oq->host_read_idx;
+ rx_bytes = 0;
+ desc_used = 0;
+ for (pkt = 0; pkt < pkts_to_process; pkt++) {
+ buff_info = (struct octep_vf_rx_buffer *)&oq->buff_info[read_idx];
+ dma_unmap_page(oq->dev, oq->desc_ring[read_idx].buffer_ptr,
+ PAGE_SIZE, DMA_FROM_DEVICE);
+ resp_hw = page_address(buff_info->page);
+ buff_info->page = NULL;
+
+ /* Swap the length field that is in Big-Endian to CPU */
+ buff_info->len = be64_to_cpu(resp_hw->length);
+ if (oct->fw_info.rx_ol_flags) {
+ /* Extended response header is immediately after
+ * response header (resp_hw)
+ */
+ resp_hw_ext = (struct octep_vf_oq_resp_hw_ext *)
+ (resp_hw + 1);
+ buff_info->len -= OCTEP_VF_OQ_RESP_HW_EXT_SIZE;
+ /* Packet Data is immediately after
+ * extended response header.
+ */
+ data_offset = OCTEP_VF_OQ_RESP_HW_SIZE +
+ OCTEP_VF_OQ_RESP_HW_EXT_SIZE;
+ rx_ol_flags = resp_hw_ext->rx_ol_flags;
+ } else {
+ /* Data is immediately after
+ * Hardware Rx response header.
+ */
+ data_offset = OCTEP_VF_OQ_RESP_HW_SIZE;
+ rx_ol_flags = 0;
+ }
+ rx_bytes += buff_info->len;
+
+ if (buff_info->len <= oq->max_single_buffer_size) {
+ skb = build_skb((void *)resp_hw, PAGE_SIZE);
+ skb_reserve(skb, data_offset);
+ skb_put(skb, buff_info->len);
+ read_idx++;
+ desc_used++;
+ if (read_idx == oq->max_count)
+ read_idx = 0;
+ } else {
+ struct skb_shared_info *shinfo;
+ u16 data_len;
+
+ skb = build_skb((void *)resp_hw, PAGE_SIZE);
+ skb_reserve(skb, data_offset);
+ /* Head fragment includes response header(s);
+ * subsequent fragments contains only data.
+ */
+ skb_put(skb, oq->max_single_buffer_size);
+ read_idx++;
+ desc_used++;
+ if (read_idx == oq->max_count)
+ read_idx = 0;
+
+ shinfo = skb_shinfo(skb);
+ data_len = buff_info->len - oq->max_single_buffer_size;
+ while (data_len) {
+ dma_unmap_page(oq->dev, oq->desc_ring[read_idx].buffer_ptr,
+ PAGE_SIZE, DMA_FROM_DEVICE);
+ buff_info = (struct octep_vf_rx_buffer *)
+ &oq->buff_info[read_idx];
+ if (data_len < oq->buffer_size) {
+ buff_info->len = data_len;
+ data_len = 0;
+ } else {
+ buff_info->len = oq->buffer_size;
+ data_len -= oq->buffer_size;
+ }
+
+ skb_add_rx_frag(skb, shinfo->nr_frags,
+ buff_info->page, 0,
+ buff_info->len,
+ buff_info->len);
+ buff_info->page = NULL;
+ read_idx++;
+ desc_used++;
+ if (read_idx == oq->max_count)
+ read_idx = 0;
+ }
+ }
+
+ skb->dev = oq->netdev;
+ skb->protocol = eth_type_trans(skb, skb->dev);
+ if (feat & NETIF_F_RXCSUM &&
+ OCTEP_VF_RX_CSUM_VERIFIED(rx_ol_flags))
+ skb->ip_summed = CHECKSUM_UNNECESSARY;
+ else
+ skb->ip_summed = CHECKSUM_NONE;
+ napi_gro_receive(oq->napi, skb);
+ }
+
+ oq->host_read_idx = read_idx;
+ oq->refill_count += desc_used;
+ oq->stats.packets += pkt;
+ oq->stats.bytes += rx_bytes;
+
+ return pkt;
+}
+
+/**
+ * octep_vf_oq_process_rx() - Process Rx queue.
+ *
+ * @oq: Octeon Rx queue data structure.
+ * @budget: max number of packets can be processed in one invocation.
+ *
+ * Check for newly received packets and process them.
+ * Keeps checking for new packets until budget is used or no new packets seen.
+ *
+ * Return: number of packets processed.
+ */
+int octep_vf_oq_process_rx(struct octep_vf_oq *oq, int budget)
+{
+ u32 pkts_available, pkts_processed, total_pkts_processed;
+ struct octep_vf_device *oct = oq->octep_vf_dev;
+
+ pkts_available = 0;
+ pkts_processed = 0;
+ total_pkts_processed = 0;
+ while (total_pkts_processed < budget) {
+ /* update pending count only when current one exhausted */
+ if (oq->pkts_pending == 0)
+ octep_vf_oq_check_hw_for_pkts(oct, oq);
+ pkts_available = min(budget - total_pkts_processed,
+ oq->pkts_pending);
+ if (!pkts_available)
+ break;
+
+ pkts_processed = __octep_vf_oq_process_rx(oct, oq,
+ pkts_available);
+ oq->pkts_pending -= pkts_processed;
+ total_pkts_processed += pkts_processed;
+ }
+
+ if (oq->refill_count >= oq->refill_threshold) {
+ u32 desc_refilled = octep_vf_oq_refill(oct, oq);
+
+ /* flush pending writes before updating credits */
+ smp_wmb();
+ writel(desc_refilled, oq->pkts_credit_reg);
+ }
+
+ return total_pkts_processed;
+}
@@ -23,6 +23,78 @@ static void octep_vf_iq_reset_indices(struct octep_vf_iq *iq)
iq->pkt_in_done = 0;
}
+/**
+ * octep_vf_iq_process_completions() - Process Tx queue completions.
+ *
+ * @iq: Octeon Tx queue data structure.
+ * @budget: max number of completions to be processed in one invocation.
+ */
+int octep_vf_iq_process_completions(struct octep_vf_iq *iq, u16 budget)
+{
+ u32 compl_pkts, compl_bytes, compl_sg;
+ struct octep_vf_device *oct = iq->octep_vf_dev;
+ struct octep_vf_tx_buffer *tx_buffer;
+ struct skb_shared_info *shinfo;
+ u32 fi = iq->flush_index;
+ struct sk_buff *skb;
+ u8 frags, i;
+
+ compl_pkts = 0;
+ compl_sg = 0;
+ compl_bytes = 0;
+ iq->octep_vf_read_index = oct->hw_ops.update_iq_read_idx(iq);
+
+ while (likely(budget && (fi != iq->octep_vf_read_index))) {
+ tx_buffer = iq->buff_info + fi;
+ skb = tx_buffer->skb;
+
+ fi++;
+ if (unlikely(fi == iq->max_count))
+ fi = 0;
+ compl_bytes += skb->len;
+ compl_pkts++;
+ budget--;
+
+ if (!tx_buffer->gather) {
+ dma_unmap_single(iq->dev, tx_buffer->dma,
+ tx_buffer->skb->len, DMA_TO_DEVICE);
+ dev_kfree_skb_any(skb);
+ continue;
+ }
+
+ /* Scatter/Gather */
+ shinfo = skb_shinfo(skb);
+ frags = shinfo->nr_frags;
+ compl_sg++;
+
+ dma_unmap_single(iq->dev, tx_buffer->sglist[0].dma_ptr[0],
+ tx_buffer->sglist[0].len[3], DMA_TO_DEVICE);
+
+ i = 1; /* entry 0 is main skb, unmapped above */
+ while (frags--) {
+ dma_unmap_page(iq->dev, tx_buffer->sglist[i >> 2].dma_ptr[i & 3],
+ tx_buffer->sglist[i >> 2].len[3 - (i & 3)], DMA_TO_DEVICE);
+ i++;
+ }
+
+ dev_kfree_skb_any(skb);
+ }
+
+ iq->pkts_processed += compl_pkts;
+ iq->stats.instr_completed += compl_pkts;
+ iq->stats.bytes_sent += compl_bytes;
+ iq->stats.sgentry_sent += compl_sg;
+ iq->flush_index = fi;
+
+ netdev_tx_completed_queue(iq->netdev_q, compl_pkts, compl_bytes);
+
+ if (unlikely(__netif_subqueue_stopped(iq->netdev, iq->q_no)) &&
+ (IQ_INSTR_SPACE(iq) >
+ OCTEP_VF_WAKE_QUEUE_THRESHOLD))
+ netif_wake_subqueue(iq->netdev, iq->q_no);
+ return !budget;
+}
+
/**
* octep_vf_iq_free_pending() - Free Tx buffers for pending completions.
*
@@ -195,8 +267,7 @@ static void octep_vf_free_iq(struct octep_vf_iq *iq)
desc_ring_size = OCTEP_VF_IQ_DESC_SIZE * CFG_GET_IQ_NUM_DESC(oct->conf);
- if (iq->buff_info)
- vfree(iq->buff_info);
+ vfree(iq->buff_info);
if (iq->desc_ring)
dma_free_coherent(iq->dev, desc_ring_size,