uhd-dpdk: Add DPDK-based sockets-like library

This library makes available a userspace network stack with a
socket-like interface for applications (except the sockets pass around
pointers to buffers and use the buffers directly--It's sockets + a
put/get for buffer management). Supported services are ARP and UDP.
Destinations can be unicast or broadcast. Multicast is not currently
supported.

The implementation has two driver layers. The upper layer runs within
the caller's context. The caller will make requests through lockless
ring buffers (including socket creation and packet transmission), and
the lower layer will implement the requests and provide a response.

Currently, the lower layer runs in a separate I/O thread, and the caller
will block until it receives a response.

The I/O thread's main body is in src/uhd_dpdk_driver.c. You'll find that
all I/O thread functions are prefixed by an underscore, and user thread
functions do not.

src/uhd_dpdk.c is used to initialize uhd-dpdk and bring up the network
interfaces.

src/uhd_dpdk_fops.c and src/uhd_dpdk_udp.c are for network services.

The test is a benchmark of a flow control loop using a certain made-up
protocol with credits and sequence number tracking.

1 files changed, 306 insertions, 0 deletions

diff --git a/host/lib/transport/uhd-dpdk/uhd_dpdk_fops.c b/host/lib/transport/uhd-dpdk/uhd_dpdk_fops.c
new file mode 100644
index 000000000..3acc3d709
--- /dev/null
+++ b/host/lib/transport/uhd-dpdk/uhd_dpdk_fops.c
@@ -0,0 +1,306 @@
+//
+// Copyright 2018 Ettus Research, a National Instruments Company
+//
+// SPDX-License-Identifier: GPL-3.0-or-later
+//
+#include "uhd_dpdk_fops.h"
+#include "uhd_dpdk_udp.h"
+#include <rte_malloc.h>
+#include <rte_ip.h>
+
+/************************************************
+ * I/O thread ONLY
+ *
+ * TODO: Decide whether to allow blocking on mutex
+ *     This would cause the I/O thread to sleep, which isn't desireable
+ *     Could throw in a "request completion cleanup" section in I/O thread's
+ *     main loop, though. Just keep trying until the requesting thred is woken
+ *     up. This would be to handle the case where the thread hadn't finished
+ *     setting itself up to wait on the condition variable, but the I/O thread
+ *     still got the request.
+ */
+int _uhd_dpdk_config_req_compl(struct uhd_dpdk_config_req *req, int retval)
+{
+    req->retval = retval;
+    int stat = pthread_mutex_trylock(&req->mutex);
+    if (stat) {
+        RTE_LOG(ERR, USER1, "%s: Could not lock req mutex\n", __func__);
+        return stat;
+    }
+    stat = pthread_cond_signal(&req->cond);
+    pthread_mutex_unlock(&req->mutex);
+    if (stat) {
+        RTE_LOG(ERR, USER1, "%s: Could not signal req cond\n", __func__);
+        return stat;
+    }
+    return 0;
+}
+
+int _uhd_dpdk_sock_setup(struct uhd_dpdk_config_req *req)
+{
+    int stat = 0;
+    switch (req->sock_type) {
+    case UHD_DPDK_SOCK_UDP:
+        stat = _uhd_dpdk_udp_setup(req);
+        break;
+    default:
+        stat = -EINVAL;
+        _uhd_dpdk_config_req_compl(req, -EINVAL);
+    }
+    return stat;
+}
+
+int _uhd_dpdk_sock_release(struct uhd_dpdk_config_req *req)
+{
+    int stat = 0;
+    switch (req->sock_type) {
+    case UHD_DPDK_SOCK_UDP:
+        stat = _uhd_dpdk_udp_release(req);
+        break;
+    default:
+        stat = -EINVAL;
+        _uhd_dpdk_config_req_compl(req, -EINVAL);
+    }
+
+    return stat;
+}
+
+/************************************************
+ * API calls
+ */
+struct uhd_dpdk_socket* uhd_dpdk_sock_open(unsigned int portid,
+                                           enum uhd_dpdk_sock_type t, void *sockarg)
+{
+    if (!ctx || (t >= UHD_DPDK_SOCK_TYPE_COUNT)) {
+        return NULL;
+    }
+
+    struct uhd_dpdk_port *port = find_port(portid);
+    if (!port) {
+        return NULL;
+    }
+
+    if (!port->ipv4_addr) {
+        RTE_LOG(WARNING, EAL, "Please set IPv4 address for port %u before opening socket\n", portid);
+        return NULL;
+    }
+
+
+    struct uhd_dpdk_config_req *req = (struct uhd_dpdk_config_req *) rte_zmalloc(NULL, sizeof(*req), 0);
+    if (!req) {
+        return NULL;
+    }
+
+    struct uhd_dpdk_socket *s = (struct uhd_dpdk_socket *) rte_zmalloc(NULL, sizeof(*s), 0);
+    if (!s) {
+        goto sock_open_end;
+    }
+
+    s->port = port;
+    req->sock = s;
+    req->req_type = UHD_DPDK_SOCK_OPEN;
+    req->sock_type = t;
+    req->retval = -ETIMEDOUT;
+    pthread_mutex_init(&req->mutex, NULL);
+    pthread_condattr_t condattr;
+    pthread_condattr_init(&condattr);
+    pthread_condattr_setclock(&condattr, CLOCK_MONOTONIC);
+    pthread_cond_init(&req->cond, &condattr);
+
+    switch (t) {
+    case UHD_DPDK_SOCK_UDP:
+        uhd_dpdk_udp_open(req, sockarg);
+        break;
+    default:
+        break;
+    }
+
+    if (req->retval) {
+        rte_free(s);
+        s = NULL;
+    }
+
+sock_open_end:
+    rte_free(req);
+    return s;
+}
+
+int uhd_dpdk_sock_close(struct uhd_dpdk_socket *sock)
+{
+    if (!ctx || !sock)
+        return -EINVAL;
+
+    struct uhd_dpdk_config_req *req = (struct uhd_dpdk_config_req *) rte_zmalloc(NULL, sizeof(*req), 0);
+    if (!req)
+        return -ENOMEM;
+    req->sock = sock;
+    req->req_type = UHD_DPDK_SOCK_CLOSE;
+    req->sock_type = sock->sock_type;
+    req->retval = -ETIMEDOUT;
+    pthread_mutex_init(&req->mutex, NULL);
+    pthread_cond_init(&req->cond, NULL);
+
+    switch (sock->sock_type) {
+    case UHD_DPDK_SOCK_UDP:
+        uhd_dpdk_udp_close(req);
+        break;
+    default:
+        break;
+    }
+
+    if (req->retval) {
+        rte_free(req);
+        return req->retval;
+    }
+
+    rte_free(sock);
+    return 0;
+}
+
+/*
+ * TODO:
+ *     Add blocking calls with timeout
+ *     Implementation would involve a condition variable, like config reqs
+ *     Also would create a cleanup section in I/O main loop (like config reqs)
+ */
+int uhd_dpdk_request_tx_bufs(struct uhd_dpdk_socket *sock, struct rte_mbuf **bufs,
+                             unsigned int num_bufs)
+{
+    if (!sock || !bufs || !num_bufs) {
+        return -EINVAL;
+    }
+    *bufs = NULL;
+
+    if (!sock->tx_ring)
+        return -EINVAL;
+
+    unsigned int num_tx = rte_ring_count(sock->rx_ring);
+    num_tx = (num_tx < num_bufs) ? num_tx : num_bufs;
+    if (rte_ring_dequeue_bulk(sock->rx_ring, (void **) bufs, num_tx, NULL) == 0)
+        return -ENOENT;
+    sock->tx_buf_count += num_tx;
+    return num_tx;
+}
+
+int uhd_dpdk_send(struct uhd_dpdk_socket *sock, struct rte_mbuf **bufs,
+                  unsigned int num_bufs)
+{
+    if (!sock || !bufs || !num_bufs)
+        return -EINVAL;
+    if (!sock->tx_ring)
+        return -EINVAL;
+    unsigned int num_tx = rte_ring_free_count(sock->tx_ring);
+    num_tx = (num_tx < num_bufs) ? num_tx : num_bufs;
+    switch (sock->sock_type) {
+    case UHD_DPDK_SOCK_UDP:
+        for (unsigned int i = 0; i < num_tx; i++) {
+            uhd_dpdk_udp_prep(sock, bufs[i]);
+        }
+        break;
+    default:
+        RTE_LOG(ERR, USER1, "%s: Unsupported sock type\n", __func__);
+        return -EINVAL;
+    }
+    int status = rte_ring_enqueue_bulk(sock->tx_ring, (void **) bufs, num_tx, NULL);
+    if (status == 0) {
+        RTE_LOG(ERR, USER1, "Invalid shared usage of TX ring detected\n");
+        return status;
+    }
+    sock->tx_buf_count -= num_tx;
+    return num_tx;
+}
+
+/*
+ * TODO:
+ *     Add blocking calls with timeout
+ */
+int uhd_dpdk_recv(struct uhd_dpdk_socket *sock, struct rte_mbuf **bufs,
+                  unsigned int num_bufs, unsigned int timeout)
+{
+    if (!sock || !bufs || !num_bufs)
+        return -EINVAL;
+    if (!sock->rx_ring)
+        return -EINVAL;
+    unsigned int num_rx = rte_ring_count(sock->rx_ring);
+    num_rx = (num_rx < num_bufs) ? num_rx : num_bufs;
+    if (num_rx) {
+        unsigned int avail = 0;
+        unsigned int status = rte_ring_dequeue_bulk(sock->rx_ring,
+                                    (void **) bufs, num_rx, &avail);
+        if (status == 0) {
+            RTE_LOG(ERR, USER1, "Invalid shared usage of RX ring detected\n");
+            RTE_LOG(ERR, USER1, "Requested %u, but %u available\n",
+                                 num_rx, avail);
+            return -ENOENT;
+        }
+    }
+    return num_rx;
+}
+
+void uhd_dpdk_free_buf(struct rte_mbuf *buf)
+{
+    rte_pktmbuf_free(buf);
+}
+
+void * uhd_dpdk_buf_to_data(struct uhd_dpdk_socket *sock, struct rte_mbuf *buf)
+{
+    if (!sock || !buf)
+        return NULL;
+
+    /* TODO: Support for more types? */
+    switch (sock->sock_type) {
+    case UHD_DPDK_SOCK_UDP:
+        return rte_pktmbuf_mtod_offset(buf, void *, sizeof(struct ether_hdr) +
+                                                    sizeof(struct ipv4_hdr) +
+                                                    sizeof(struct udp_hdr));
+    default:
+        return NULL;
+    }
+}
+
+
+int uhd_dpdk_get_len(struct uhd_dpdk_socket *sock, struct rte_mbuf *buf)
+{
+    if (!sock || !buf)
+        return -EINVAL;
+
+    if (sock->sock_type != UHD_DPDK_SOCK_UDP)
+        return -EINVAL;
+
+    struct udp_hdr *hdr = (struct udp_hdr *) ((uint8_t *) uhd_dpdk_buf_to_data(sock, buf) - sizeof(struct udp_hdr));
+    if (!hdr)
+        return -EINVAL;
+
+    /* Report dgram length - header */
+    return ntohs(hdr->dgram_len) - 8;
+}
+
+int uhd_dpdk_get_src_ipv4(struct uhd_dpdk_socket *sock, struct rte_mbuf *buf,
+                          uint32_t *ipv4_addr)
+{
+    if (!sock || !buf || !ipv4_addr)
+        return -EINVAL;
+
+    if (sock->sock_type != UHD_DPDK_SOCK_UDP)
+        return -EINVAL;
+
+    struct ipv4_hdr *hdr = rte_pktmbuf_mtod_offset(buf, struct ipv4_hdr *,
+                                                   sizeof(struct ether_hdr));
+
+    *ipv4_addr = hdr->src_addr;
+    return 0;
+}
+
+int uhd_dpdk_get_drop_count(struct uhd_dpdk_socket *sock, uint32_t *count)
+{
+    if (!sock)
+        return -EINVAL;
+    if (sock->sock_type != UHD_DPDK_SOCK_UDP)
+        return -EINVAL;
+    if (!sock->priv)
+        return -ENODEV;
+
+    struct uhd_dpdk_udp_priv *pdata = (struct uhd_dpdk_udp_priv *) sock->priv;
+    *count = pdata->dropped_pkts;
+    return 0;
+}