devices: add multi-queue support for af-packet
Type: feature
Change-Id: I0f4e6517fcfa07ffb0aba89b159ac1337937a508
Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
diff --git a/src/vnet/devices/af_packet/device.c b/src/vnet/devices/af_packet/device.c
index f7f006e..013d9f7 100644
--- a/src/vnet/devices/af_packet/device.c
+++ b/src/vnet/devices/af_packet/device.c
@@ -60,6 +60,7 @@
{
u32 buffer_index;
u32 hw_if_index;
+ u16 queue_id;
tpacket3_hdr_t tph;
vnet_virtio_net_hdr_t vnet_hdr;
vlib_buffer_t buffer;
@@ -87,51 +88,74 @@
af_packet_main_t *apm = &af_packet_main;
af_packet_if_t *apif = pool_elt_at_index (apm->interfaces, dev_instance);
- clib_spinlock_lock_if_init (&apif->lockp);
- u32 tx_block_sz = apif->tx_req->tp_block_size;
- u32 tx_frame_sz = apif->tx_req->tp_frame_size;
- u32 tx_frame_nr = apif->tx_req->tp_frame_nr;
- u32 tx_block_nr = apif->tx_req->tp_block_nr;
- u32 rx_block_size = apif->rx_req->tp_block_size;
- u32 rx_frame_size = apif->rx_req->tp_frame_size;
- u32 rx_frame_nr = apif->rx_req->tp_frame_nr;
- u32 rx_block_nr = apif->rx_req->tp_block_nr;
- int block = 0;
- u8 *tx_block_start = apif->tx_ring[block];
- u32 tx_frame = apif->next_tx_frame;
- tpacket3_hdr_t *tph;
+ af_packet_queue_t *rx_queue = 0;
+ af_packet_queue_t *tx_queue = 0;
- s = format (s, "Linux PACKET socket interface\n");
- s = format (s, "%UTX block size:%d nr:%d TX frame size:%d nr:%d\n",
- format_white_space, indent, tx_block_sz, tx_block_nr,
- tx_frame_sz, tx_frame_nr);
- s = format (s, "%URX block size:%d nr:%d RX frame size:%d nr:%d\n",
- format_white_space, indent, rx_block_size, rx_block_nr,
- rx_frame_size, rx_frame_nr);
- s = format (s, "%Unext frame:%d\n", format_white_space, indent,
- apif->next_tx_frame);
+ s = format (s, "Linux PACKET socket interface");
- int n_send_req = 0, n_avail = 0, n_sending = 0, n_tot = 0, n_wrong = 0;
- do
+ vec_foreach (rx_queue, apif->rx_queues)
{
- tph = (tpacket3_hdr_t *) (tx_block_start + tx_frame * tx_frame_sz);
- tx_frame = (tx_frame + 1) % tx_frame_nr;
- if (tph->tp_status == 0)
- n_avail++;
- else if (tph->tp_status & TP_STATUS_SEND_REQUEST)
- n_send_req++;
- else if (tph->tp_status & TP_STATUS_SENDING)
- n_sending++;
- else
- n_wrong++;
- n_tot++;
- }
- while (tx_frame != apif->next_tx_frame);
- s = format (s, "%Uavailable:%d request:%d sending:%d wrong:%d total:%d\n",
- format_white_space, indent, n_avail, n_send_req, n_sending,
- n_wrong, n_tot);
+ u32 rx_block_size = rx_queue->rx_req->tp_block_size;
+ u32 rx_frame_size = rx_queue->rx_req->tp_frame_size;
+ u32 rx_frame_nr = rx_queue->rx_req->tp_frame_nr;
+ u32 rx_block_nr = rx_queue->rx_req->tp_block_nr;
- clib_spinlock_unlock_if_init (&apif->lockp);
+ s = format (s, "\n%URX Queue %u:", format_white_space, indent,
+ rx_queue->queue_id);
+ s = format (s, "\n%Ublock size:%d nr:%d frame size:%d nr:%d",
+ format_white_space, indent + 2, rx_block_size, rx_block_nr,
+ rx_frame_size, rx_frame_nr);
+ s = format (s, " next block:%d", rx_queue->next_rx_block);
+ if (rx_queue->is_rx_pending)
+ {
+ s = format (
+ s, "\n%UPending Request: num-rx-pkts:%d next-frame-offset:%d",
+ format_white_space, indent + 2, rx_queue->num_rx_pkts,
+ rx_queue->rx_frame_offset);
+ }
+ }
+
+ vec_foreach (tx_queue, apif->tx_queues)
+ {
+ clib_spinlock_lock (&tx_queue->lockp);
+ u32 tx_block_sz = tx_queue->tx_req->tp_block_size;
+ u32 tx_frame_sz = tx_queue->tx_req->tp_frame_size;
+ u32 tx_frame_nr = tx_queue->tx_req->tp_frame_nr;
+ u32 tx_block_nr = tx_queue->tx_req->tp_block_nr;
+ int block = 0;
+ int n_send_req = 0, n_avail = 0, n_sending = 0, n_tot = 0, n_wrong = 0;
+ u8 *tx_block_start = tx_queue->tx_ring[block];
+ u32 tx_frame = tx_queue->next_tx_frame;
+ tpacket3_hdr_t *tph;
+
+ s = format (s, "\n%UTX Queue %u:", format_white_space, indent,
+ tx_queue->queue_id);
+ s = format (s, "\n%Ublock size:%d nr:%d frame size:%d nr:%d",
+ format_white_space, indent + 2, tx_block_sz, tx_block_nr,
+ tx_frame_sz, tx_frame_nr);
+ s = format (s, " next frame:%d", tx_queue->next_tx_frame);
+
+ do
+ {
+ tph = (tpacket3_hdr_t *) (tx_block_start + tx_frame * tx_frame_sz);
+ tx_frame = (tx_frame + 1) % tx_frame_nr;
+ if (tph->tp_status == 0)
+ n_avail++;
+ else if (tph->tp_status & TP_STATUS_SEND_REQUEST)
+ n_send_req++;
+ else if (tph->tp_status & TP_STATUS_SENDING)
+ n_sending++;
+ else
+ n_wrong++;
+ n_tot++;
+ }
+ while (tx_frame != tx_queue->next_tx_frame);
+ s =
+ format (s, "\n%Uavailable:%d request:%d sending:%d wrong:%d total:%d",
+ format_white_space, indent + 2, n_avail, n_send_req, n_sending,
+ n_wrong, n_tot);
+ clib_spinlock_unlock (&tx_queue->lockp);
+ }
return s;
}
@@ -143,7 +167,8 @@
af_packet_tx_trace_t *t = va_arg (*va, af_packet_tx_trace_t *);
u32 indent = format_get_indent (s);
- s = format (s, "af_packet: hw_if_index %u", t->hw_if_index);
+ s = format (s, "af_packet: hw_if_index %u tx-queue %u", t->hw_if_index,
+ t->queue_id);
s =
format (s,
@@ -183,11 +208,13 @@
static void
af_packet_tx_trace (vlib_main_t *vm, vlib_node_runtime_t *node,
vlib_buffer_t *b0, u32 bi, tpacket3_hdr_t *tph,
- vnet_virtio_net_hdr_t *vnet_hdr, u32 hw_if_index)
+ vnet_virtio_net_hdr_t *vnet_hdr, u32 hw_if_index,
+ u16 queue_id)
{
af_packet_tx_trace_t *t;
t = vlib_add_trace (vm, node, b0, sizeof (t[0]));
t->hw_if_index = hw_if_index;
+ t->queue_id = queue_id;
t->buffer_index = bi;
clib_memcpy_fast (&t->tph, tph, sizeof (*tph));
@@ -289,22 +316,29 @@
vlib_frame_t * frame)
{
af_packet_main_t *apm = &af_packet_main;
+ vnet_hw_if_tx_frame_t *tf = vlib_frame_scalar_args (frame);
u32 *buffers = vlib_frame_vector_args (frame);
u32 n_left = frame->n_vectors;
u32 n_sent = 0;
vnet_interface_output_runtime_t *rd = (void *) node->runtime_data;
af_packet_if_t *apif =
pool_elt_at_index (apm->interfaces, rd->dev_instance);
- clib_spinlock_lock_if_init (&apif->lockp);
- u32 block = 0;
- u32 frame_size = apif->tx_req->tp_frame_size;
- u32 frame_num = apif->tx_req->tp_frame_nr;
- u8 *block_start = apif->tx_ring[block];
- u32 tx_frame = apif->next_tx_frame;
- tpacket3_hdr_t *tph;
+ u16 queue_id = tf->queue_id;
+ af_packet_queue_t *tx_queue = vec_elt_at_index (apif->tx_queues, queue_id);
+ u32 block = 0, frame_size = 0, frame_num = 0, tx_frame = 0;
+ u8 *block_start = 0;
+ tpacket3_hdr_t *tph = 0;
u32 frame_not_ready = 0;
u8 is_cksum_gso_enabled = (apif->is_cksum_gso_enabled == 1) ? 1 : 0;
+ if (tf->shared_queue)
+ clib_spinlock_lock (&tx_queue->lockp);
+
+ frame_size = tx_queue->tx_req->tp_frame_size;
+ frame_num = tx_queue->tx_req->tp_frame_nr;
+ block_start = tx_queue->tx_ring[block];
+ tx_frame = tx_queue->next_tx_frame;
+
while (n_left)
{
u32 len;
@@ -366,12 +400,12 @@
{
if (PREDICT_TRUE (is_cksum_gso_enabled))
af_packet_tx_trace (vm, node, b0_first, bi_first, tph, vnet_hdr,
- apif->hw_if_index);
+ apif->hw_if_index, queue_id);
else
{
vnet_virtio_net_hdr_t vnet_hdr2 = {};
af_packet_tx_trace (vm, node, b0_first, bi_first, tph,
- &vnet_hdr2, apif->hw_if_index);
+ &vnet_hdr2, apif->hw_if_index, queue_id);
}
}
tx_frame = (tx_frame + 1) % frame_num;
@@ -386,10 +420,10 @@
if (PREDICT_TRUE (n_sent))
{
- apif->next_tx_frame = tx_frame;
+ tx_queue->next_tx_frame = tx_frame;
- if (PREDICT_FALSE (sendto (apif->fd, NULL, 0, MSG_DONTWAIT, NULL, 0) ==
- -1))
+ if (PREDICT_FALSE (
+ sendto (tx_queue->fd, NULL, 0, MSG_DONTWAIT, NULL, 0) == -1))
{
/* Uh-oh, drop & move on, but count whether it was fatal or not.
* Note that we have no reliable way to properly determine the
@@ -403,7 +437,8 @@
}
}
- clib_spinlock_unlock_if_init (&apif->lockp);
+ if (tf->shared_queue)
+ clib_spinlock_unlock (&tx_queue->lockp);
if (PREDICT_FALSE (frame_not_ready))
vlib_error_count (vm, node->node_index,