* [PATCH] net/af_xdp: AF_XDP PMD CNI Integration
@ 2022-12-13 10:47 Koikkara Reeny, Shibin
2023-01-18 12:11 ` Bruce Richardson
0 siblings, 1 reply; 2+ messages in thread
From: Koikkara Reeny, Shibin @ 2022-12-13 10:47 UTC (permalink / raw)
To: dev, anatoly.burakov; +Cc: ciara.loftus, qi.z.zhang, Shibin Koikkara Reeny
From: Shibin Koikkara Reeny <shibin.koikkara.reeny@intel.com>
Integrate support for the AF_XDP CNI and device plugin so that the
DPDK AF_XDP PMD can work in an unprivileged container environment.
Part of the AF_XDP PMD initialization process involves loading
an eBPF program onto the given netdev. This operation requires
privileges, which prevents the PMD from being able to work in an
unprivileged container (without root access). The plugin CNI handles
the program loading. CNI open Unix Domain Socket (UDS) and waits
listening for a client to make requests over that UDS. The client(DPDK)
connects and a "handshake" occurs, then the File Descriptor which points
to the XSKMAP associated with the loaded eBPF program is handed over
to the client. The client can then proceed with creating an AF_XDP
socket and inserting the socket into the XSKMAP pointed to by the
FD received on the UDS.
A new vdev arg "use_cni" is created to indicate user wishes to run
the PMD in unprivileged mode and to receive the XSKMAP FD from the CNI.
When this flag is set, the XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD libbpf flag
should be used when creating the socket, which tells libbpf not to load the
default libbpf program on the netdev. We tell libbpf not to do this because
the loading is handled by the CNI in this scenario.
Signed-off-by: Shibin Koikkara Reeny <shibin.koikkara.reeny@intel.com>
---
drivers/net/af_xdp/rte_eth_af_xdp.c | 337 +++++++++++++++++++++++++++-
1 file changed, 325 insertions(+), 12 deletions(-)
diff --git a/drivers/net/af_xdp/rte_eth_af_xdp.c b/drivers/net/af_xdp/rte_eth_af_xdp.c
index b6ec9bf490..b69700a0c7 100644
--- a/drivers/net/af_xdp/rte_eth_af_xdp.c
+++ b/drivers/net/af_xdp/rte_eth_af_xdp.c
@@ -7,6 +7,7 @@
#include <string.h>
#include <netinet/in.h>
#include <net/if.h>
+#include <sys/un.h>
#include <sys/socket.h>
#include <sys/ioctl.h>
#include <linux/if_ether.h>
@@ -81,6 +82,24 @@ RTE_LOG_REGISTER_DEFAULT(af_xdp_logtype, NOTICE);
#define ETH_AF_XDP_MP_KEY "afxdp_mp_send_fds"
+#define MAX_LONG_OPT_SZ 64
+#define UDS_MAX_FD_NUM 2
+#define UDS_MAX_CMD_LEN 64
+#define UDS_MAX_CMD_RESP 128
+#define UDS_XSK_MAP_FD_MSG "/xsk_map_fd"
+#define UDS_SOCK "/tmp/afxdp.sock"
+#define UDS_CONNECT_MSG "/connect"
+#define UDS_HOST_OK_MSG "/host_ok"
+#define UDS_HOST_NAK_MSG "/host_nak"
+#define UDS_VERSION_MSG "/version"
+#define UDS_XSK_MAP_FD_MSG "/xsk_map_fd"
+#define UDS_XSK_SOCKET_MSG "/xsk_socket"
+#define UDS_FD_ACK_MSG "/fd_ack"
+#define UDS_FD_NAK_MSG "/fd_nak"
+#define UDS_FIN_MSG "/fin"
+#define UDS_FIN_ACK_MSG "/fin_ack"
+
+
static int afxdp_dev_count;
/* Message header to synchronize fds via IPC */
@@ -151,6 +170,7 @@ struct pmd_internals {
char prog_path[PATH_MAX];
bool custom_prog_configured;
bool force_copy;
+ bool use_cni;
struct bpf_map *map;
struct rte_ether_addr eth_addr;
@@ -170,6 +190,7 @@ struct pmd_process_private {
#define ETH_AF_XDP_PROG_ARG "xdp_prog"
#define ETH_AF_XDP_BUDGET_ARG "busy_budget"
#define ETH_AF_XDP_FORCE_COPY_ARG "force_copy"
+#define ETH_AF_XDP_USE_CNI_ARG "use_cni"
static const char * const valid_arguments[] = {
ETH_AF_XDP_IFACE_ARG,
@@ -179,8 +200,8 @@ static const char * const valid_arguments[] = {
ETH_AF_XDP_PROG_ARG,
ETH_AF_XDP_BUDGET_ARG,
ETH_AF_XDP_FORCE_COPY_ARG,
- NULL
-};
+ ETH_AF_XDP_USE_CNI_ARG,
+ NULL};
static const struct rte_eth_link pmd_link = {
.link_speed = RTE_ETH_SPEED_NUM_10G,
@@ -1129,7 +1150,8 @@ xsk_umem_info *xdp_umem_configure(struct pmd_internals *internals,
ret = xsk_umem__create(&umem->umem, base_addr, umem_size,
&rxq->fq, &rxq->cq, &usr_config);
if (ret) {
- AF_XDP_LOG(ERR, "Failed to create umem\n");
+ AF_XDP_LOG(ERR, "Failed to create umem [%d]: [%s]\n",
+ errno, strerror(errno));
goto err;
}
umem->buffer = base_addr;
@@ -1314,6 +1336,245 @@ configure_preferred_busy_poll(struct pkt_rx_queue *rxq)
return 0;
}
+static int
+init_uds_sock(struct sockaddr_un *server)
+{
+ int sock;
+
+ sock = socket(AF_UNIX, SOCK_SEQPACKET, 0);
+ if (sock < 0) {
+ AF_XDP_LOG(ERR, "Failed to opening stream socket\n");
+ return -1;
+ }
+
+ server->sun_family = AF_UNIX;
+ strlcpy(server->sun_path, UDS_SOCK, sizeof(server->sun_path));
+
+ if (connect(sock, (struct sockaddr *)server, sizeof(struct sockaddr_un)) < 0) {
+ close(sock);
+ AF_XDP_LOG(ERR, "Error connecting stream socket errno = [%d]: [%s]\n",
+ errno, strerror(errno));
+ return -1;
+ }
+
+ return sock;
+}
+
+struct msg_internal {
+ char response[UDS_MAX_CMD_RESP];
+ int len_param;
+ int num_fds;
+ int fds[UDS_MAX_FD_NUM];
+};
+
+static int
+send_msg(int sock, char *request, int *fd)
+{
+ int snd;
+ struct iovec iov;
+ struct msghdr msgh;
+ struct cmsghdr *cmsg;
+ struct sockaddr_un dst;
+ char control[CMSG_SPACE(sizeof(*fd))];
+
+ memset(&dst, 0, sizeof(dst));
+ dst.sun_family = AF_UNIX;
+ strlcpy(dst.sun_path, UDS_SOCK, sizeof(dst.sun_path));
+
+ /* Initialize message header structure */
+ memset(&msgh, 0, sizeof(msgh));
+ memset(control, 0, sizeof(control));
+ iov.iov_base = request;
+ iov.iov_len = strlen(request);
+
+ msgh.msg_name = &dst;
+ msgh.msg_namelen = sizeof(dst);
+ msgh.msg_iov = &iov;
+ msgh.msg_iovlen = 1;
+ msgh.msg_control = control;
+ msgh.msg_controllen = sizeof(control);
+
+ /* Translate the FD. */
+ cmsg = CMSG_FIRSTHDR(&msgh);
+ cmsg->cmsg_len = CMSG_LEN(sizeof(*fd));
+ cmsg->cmsg_level = SOL_SOCKET;
+ cmsg->cmsg_type = SCM_RIGHTS;
+ memcpy(CMSG_DATA(cmsg), fd, sizeof(*fd));
+
+ /* Send the request message. */
+ do {
+ snd = sendmsg(sock, &msgh, 0);
+ } while (snd < 0 && errno == EINTR);
+
+ return snd;
+}
+
+static int
+read_msg(int sock, char *response, struct sockaddr_un *s, int *fd)
+{
+ int msglen;
+ struct msghdr msgh;
+ struct iovec iov;
+ char control[CMSG_SPACE(sizeof(*fd))];
+ struct cmsghdr *cmsg;
+
+ /* Initialize message header structure */
+ memset(&msgh, 0, sizeof(msgh));
+ iov.iov_base = response;
+ iov.iov_len = UDS_MAX_CMD_RESP;
+
+ msgh.msg_name = s;
+ msgh.msg_namelen = sizeof(*s);
+ msgh.msg_iov = &iov;
+ msgh.msg_iovlen = 1;
+ msgh.msg_control = control;
+ msgh.msg_controllen = sizeof(control);
+
+ msglen = recvmsg(sock, &msgh, 0);
+
+ /* zero length message means socket was closed */
+ if (msglen == 0)
+ return 0;
+
+ if (msglen < 0) {
+ AF_XDP_LOG(ERR, "recvmsg failed, %s\n", strerror(errno));
+ return -1;
+ }
+
+ /* read auxiliary FDs if any */
+ for (cmsg = CMSG_FIRSTHDR(&msgh); cmsg != NULL;
+ cmsg = CMSG_NXTHDR(&msgh, cmsg)) {
+ if (cmsg->cmsg_level == SOL_SOCKET &&
+ cmsg->cmsg_type == SCM_RIGHTS) {
+ memcpy(fd, CMSG_DATA(cmsg), sizeof(*fd));
+ break;
+ }
+ }
+
+ response[msglen] = '\0';
+ return msglen;
+}
+
+static int
+make_request_cni(int sock, struct sockaddr_un *server, char *request,
+ int *req_fd, char *response, int *out_fd)
+{
+ int rval;
+
+ AF_XDP_LOG(INFO, "Request: [%s]\n", request);
+
+ /* if no file descriptor to send then directly write to socket.
+ * else use sendmsg() to send the file descriptor.
+ */
+ if (req_fd == NULL)
+ rval = write(sock, request, strlen(request));
+ else
+ rval = send_msg(sock, request, req_fd);
+
+ if (rval < 0) {
+ AF_XDP_LOG(ERR, "Write error %s\n", strerror(errno));
+ return -1;
+ }
+
+ rval = read_msg(sock, response, server, out_fd);
+ if (rval <= 0) {
+ AF_XDP_LOG(ERR, "Read error %d\n", rval);
+ return -1;
+ }
+ AF_XDP_LOG(INFO, "Response: [%s]\n", request);
+
+ return 0;
+}
+
+static int
+check_response(char *response, char *exp_resp, long size)
+{
+ return strncmp(response, exp_resp, size);
+}
+
+static int
+get_cni_fd(char *if_name)
+{
+ char request[UDS_MAX_CMD_LEN], response[UDS_MAX_CMD_RESP];
+ char hostname[MAX_LONG_OPT_SZ], exp_resp[UDS_MAX_CMD_RESP];
+ struct sockaddr_un server;
+ int xsk_map_fd = -1, out_fd = 0;
+ int sock, err;
+
+ err = gethostname(hostname, MAX_LONG_OPT_SZ - 1);
+ if (err)
+ return -1;
+
+ memset(&server, 0, sizeof(server));
+ sock = init_uds_sock(&server);
+
+ /* Initiates handshake to CNI send: /connect,hostname */
+ snprintf(request, sizeof(request), "%s,%s", UDS_CONNECT_MSG, hostname);
+ memset(response, 0, sizeof(response));
+ if (make_request_cni(sock, &server, request, NULL, response, &out_fd) < 0) {
+ AF_XDP_LOG(ERR, "Error in proccessing cmd [%s]\n", request);
+ goto err_close;
+ }
+
+ /* Expect /host_ok */
+ strlcpy(exp_resp, UDS_HOST_OK_MSG, UDS_MAX_CMD_LEN);
+ if (check_response(response, exp_resp, strlen(exp_resp)) < 0) {
+ AF_XDP_LOG(ERR, "Unexpected response [%s]\n", response);
+ goto err_close;
+ }
+ /* Request for "/version" */
+ strlcpy(request, UDS_VERSION_MSG, UDS_MAX_CMD_LEN);
+ memset(response, 0, sizeof(response));
+ if (make_request_cni(sock, &server, request, NULL, response, &out_fd) < 0) {
+ AF_XDP_LOG(ERR, "Error in proccessing cmd [%s]\n", request);
+ goto err_close;
+ }
+
+ /* Request for file descriptor for netdev name*/
+ snprintf(request, sizeof(request), "%s,%s", UDS_XSK_MAP_FD_MSG, if_name);
+ memset(response, 0, sizeof(response));
+ if (make_request_cni(sock, &server, request, NULL, response, &out_fd) < 0) {
+ AF_XDP_LOG(ERR, "Error in proccessing cmd [%s]\n", request);
+ goto err_close;
+ }
+
+ if (out_fd < 0) {
+ AF_XDP_LOG(ERR, "Error in proccessing cmd [%s]\n", request);
+ goto err_close;
+ }
+
+ xsk_map_fd = out_fd;
+
+ /* Expect fd_ack with file descriptor */
+ strlcpy(exp_resp, UDS_FD_ACK_MSG, UDS_MAX_CMD_LEN);
+ if (check_response(response, exp_resp, strlen(exp_resp)) < 0) {
+ AF_XDP_LOG(ERR, "Unexpected response [%s]\n", response);
+ goto err_close;
+ }
+
+ /* Initiate close connection */
+ strlcpy(request, UDS_FIN_MSG, UDS_MAX_CMD_LEN);
+ memset(response, 0, sizeof(response));
+ if (make_request_cni(sock, &server, request, NULL, response, &out_fd) < 0) {
+ AF_XDP_LOG(ERR, "Error in proccessing cmd [%s]\n", request);
+ goto err_close;
+ }
+
+ /* Connection close */
+ strlcpy(exp_resp, UDS_FIN_ACK_MSG, UDS_MAX_CMD_LEN);
+ if (check_response(response, exp_resp, strlen(exp_resp)) < 0) {
+ AF_XDP_LOG(ERR, "Unexpected response [%s]\n", response);
+ goto err_close;
+ }
+ close(sock);
+
+ return xsk_map_fd;
+
+err_close:
+ close(sock);
+ return -1;
+}
+
static int
xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq,
int ring_size)
@@ -1362,6 +1623,10 @@ xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq,
cfg.bind_flags |= XDP_USE_NEED_WAKEUP;
#endif
+ /* Disable libbpf from loading XDP program */
+ if (internals->use_cni)
+ cfg.libbpf_flags |= XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD;
+
if (strnlen(internals->prog_path, PATH_MAX)) {
if (!internals->custom_prog_configured) {
ret = load_custom_xdp_prog(internals->prog_path,
@@ -1413,7 +1678,23 @@ xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq,
}
}
- if (rxq->busy_budget) {
+ if (internals->use_cni) {
+ int err, fd, map_fd;
+
+ /* get socket fd from CNI plugin */
+ map_fd = get_cni_fd(internals->if_name);
+ if (map_fd < 0) {
+ AF_XDP_LOG(ERR, "Failed to receive CNI plugin fd\n");
+ goto out_xsk;
+ }
+ /* get socket fd */
+ fd = xsk_socket__fd(rxq->xsk);
+ err = bpf_map_update_elem(map_fd, &rxq->xsk_queue_idx, &fd, 0);
+ if (err) {
+ AF_XDP_LOG(ERR, "Failed to insert unprivileged xsk in map.\n");
+ goto out_xsk;
+ }
+ } else if (rxq->busy_budget) {
ret = configure_preferred_busy_poll(rxq);
if (ret) {
AF_XDP_LOG(ERR, "Failed configure busy polling.\n");
@@ -1584,6 +1865,26 @@ static const struct eth_dev_ops ops = {
.get_monitor_addr = eth_get_monitor_addr,
};
+/* CNI option works in unprivileged container environment
+ * and ethernet device functionality will be reduced. So
+ * additional customiszed eth_dev_ops struct is needed
+ * for cni.
+ **/
+static const struct eth_dev_ops ops_cni = {
+ .dev_start = eth_dev_start,
+ .dev_stop = eth_dev_stop,
+ .dev_close = eth_dev_close,
+ .dev_configure = eth_dev_configure,
+ .dev_infos_get = eth_dev_info,
+ .mtu_set = eth_dev_mtu_set,
+ .rx_queue_setup = eth_rx_queue_setup,
+ .tx_queue_setup = eth_tx_queue_setup,
+ .link_update = eth_link_update,
+ .stats_get = eth_stats_get,
+ .stats_reset = eth_stats_reset,
+ .get_monitor_addr = eth_get_monitor_addr,
+};
+
/** parse busy_budget argument */
static int
parse_budget_arg(const char *key __rte_unused,
@@ -1704,8 +2005,8 @@ xdp_get_channels_info(const char *if_name, int *max_queues,
static int
parse_parameters(struct rte_kvargs *kvlist, char *if_name, int *start_queue,
- int *queue_cnt, int *shared_umem, char *prog_path,
- int *busy_budget, int *force_copy)
+ int *queue_cnt, int *shared_umem, char *prog_path,
+ int *busy_budget, int *force_copy, int *use_cni)
{
int ret;
@@ -1746,6 +2047,11 @@ parse_parameters(struct rte_kvargs *kvlist, char *if_name, int *start_queue,
if (ret < 0)
goto free_kvlist;
+ ret = rte_kvargs_process(kvlist, ETH_AF_XDP_USE_CNI_ARG,
+ &parse_integer_arg, use_cni);
+ if (ret < 0)
+ goto free_kvlist;
+
free_kvlist:
rte_kvargs_free(kvlist);
return ret;
@@ -1783,8 +2089,9 @@ get_iface_info(const char *if_name,
static struct rte_eth_dev *
init_internals(struct rte_vdev_device *dev, const char *if_name,
- int start_queue_idx, int queue_cnt, int shared_umem,
- const char *prog_path, int busy_budget, int force_copy)
+ int start_queue_idx, int queue_cnt, int shared_umem,
+ const char *prog_path, int busy_budget, int force_copy,
+ int use_cni)
{
const char *name = rte_vdev_device_name(dev);
const unsigned int numa_node = dev->device.numa_node;
@@ -1813,6 +2120,7 @@ init_internals(struct rte_vdev_device *dev, const char *if_name,
#endif
internals->shared_umem = shared_umem;
internals->force_copy = force_copy;
+ internals->use_cni = use_cni;
if (xdp_get_channels_info(if_name, &internals->max_queue_cnt,
&internals->combined_queue_cnt)) {
@@ -1871,7 +2179,11 @@ init_internals(struct rte_vdev_device *dev, const char *if_name,
eth_dev->data->dev_link = pmd_link;
eth_dev->data->mac_addrs = &internals->eth_addr;
eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
- eth_dev->dev_ops = &ops;
+ if (!internals->use_cni)
+ eth_dev->dev_ops = &ops;
+ else
+ eth_dev->dev_ops = &ops_cni;
+
eth_dev->rx_pkt_burst = eth_af_xdp_rx;
eth_dev->tx_pkt_burst = eth_af_xdp_tx;
eth_dev->process_private = process_private;
@@ -1998,6 +2310,7 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device *dev)
char prog_path[PATH_MAX] = {'\0'};
int busy_budget = -1, ret;
int force_copy = 0;
+ int use_cni = 0;
struct rte_eth_dev *eth_dev = NULL;
const char *name = rte_vdev_device_name(dev);
@@ -2043,7 +2356,7 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device *dev)
if (parse_parameters(kvlist, if_name, &xsk_start_queue_idx,
&xsk_queue_cnt, &shared_umem, prog_path,
- &busy_budget, &force_copy) < 0) {
+ &busy_budget, &force_copy, &use_cni) < 0) {
AF_XDP_LOG(ERR, "Invalid kvargs value\n");
return -EINVAL;
}
@@ -2057,8 +2370,8 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device *dev)
busy_budget;
eth_dev = init_internals(dev, if_name, xsk_start_queue_idx,
- xsk_queue_cnt, shared_umem, prog_path,
- busy_budget, force_copy);
+ xsk_queue_cnt, shared_umem, prog_path,
+ busy_budget, force_copy, use_cni);
if (eth_dev == NULL) {
AF_XDP_LOG(ERR, "Failed to init internals\n");
return -1;
--
2.25.1
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] net/af_xdp: AF_XDP PMD CNI Integration
2022-12-13 10:47 [PATCH] net/af_xdp: AF_XDP PMD CNI Integration Koikkara Reeny, Shibin
@ 2023-01-18 12:11 ` Bruce Richardson
0 siblings, 0 replies; 2+ messages in thread
From: Bruce Richardson @ 2023-01-18 12:11 UTC (permalink / raw)
To: Koikkara Reeny, Shibin; +Cc: dev, anatoly.burakov, ciara.loftus, qi.z.zhang
On Tue, Dec 13, 2022 at 10:47:46AM +0000, Koikkara Reeny, Shibin wrote:
> From: Shibin Koikkara Reeny <shibin.koikkara.reeny@intel.com>
>
> Integrate support for the AF_XDP CNI and device plugin so that the
> DPDK AF_XDP PMD can work in an unprivileged container environment.
> Part of the AF_XDP PMD initialization process involves loading
> an eBPF program onto the given netdev. This operation requires
> privileges, which prevents the PMD from being able to work in an
> unprivileged container (without root access). The plugin CNI handles
> the program loading. CNI open Unix Domain Socket (UDS) and waits
> listening for a client to make requests over that UDS. The client(DPDK)
> connects and a "handshake" occurs, then the File Descriptor which points
> to the XSKMAP associated with the loaded eBPF program is handed over
> to the client. The client can then proceed with creating an AF_XDP
> socket and inserting the socket into the XSKMAP pointed to by the
> FD received on the UDS.
>
> A new vdev arg "use_cni" is created to indicate user wishes to run
> the PMD in unprivileged mode and to receive the XSKMAP FD from the CNI.
> When this flag is set, the XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD libbpf flag
> should be used when creating the socket, which tells libbpf not to load the
> default libbpf program on the netdev. We tell libbpf not to do this because
> the loading is handled by the CNI in this scenario.
>
> Signed-off-by: Shibin Koikkara Reeny <shibin.koikkara.reeny@intel.com>
> ---
> drivers/net/af_xdp/rte_eth_af_xdp.c | 337 +++++++++++++++++++++++++++-
> 1 file changed, 325 insertions(+), 12 deletions(-)
>
Hi Shibhin,
this support sounds really interesting, but I think the patch needs to
include documentation updates on how to set up and use this feature. No
need to duplicate instructions that may be already given in the CNI/plugin
docs, but we need to at least reference those from our docs to ensure end
users can quickly find out how to work with this.
Thanks,
/Bruce
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2023-01-18 12:11 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-12-13 10:47 [PATCH] net/af_xdp: AF_XDP PMD CNI Integration Koikkara Reeny, Shibin
2023-01-18 12:11 ` Bruce Richardson
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).