Skip to contentSkip to conversation

net/af_xdp: fix multi interface support for K8s

PATCH Superseded Maryam Tahhan 2 participants 6 messages Feb 29, 2024 – Mar 1, 2024 Network drivers

v11 of 14 Superseded Latest is v14 →

lore
Share & tools
lore
#0 lore raw

v11 cover net/af_xdp: fix multi interface support for K8s

The original use_cni implementation was limited to supporting only a single netdev in a DPDK pod. This patchset aims to fix this limitation transparently to the end user. It will also enable compatibility with the latest AF_XDP Device Plugin.

---
v11:
* Fixed up typos picked up by checkpatch.

v10:
* Add UDS acronym
* Update use_cni in docs with ``use_cni``
* Remove reference to limitations and simply document behaviour before and after DPDK 23.11.

v9:
* Fixup checkpatch issues.

v8:
* Go back to using use_cni vdev argument
* Introduce use_map_pinning vdev param.
* Rename uds_path to dp_path so that it can be used with map pinning as well as use_cni.
* Set dp_path internally in the AF_XDP PMD if it's not configured by the user.
* Clean up the original use_cni documentation separately to coding changes.

v7:
* Give a more descriptive commit msg headline.
* Fixup typos in documentation.

v6:
* Add link to PR 81 in commit message
* Add release notes changes to this patchset

v5:
* Fix alignment for ETH_AF_XDP_USE_DP_UDS_PATH_ARG
* Remove use_cni references in af_xdp.rst

v4:
* Rename af_xdp_cni.rst to af_xdp_dp.rst
* Removed all incorrect references to CNI throughout af_xdp PMD file.
* Fixed Typos in af_xdp_dp.rst

v3:
* Remove use_cni vdev argument as it's no longer needed.
* Update incorrect CNI references for the AF_XDP DP in the documentation.
* Update the documentation to run a simple example with the AF_XDP DP plugin in K8s.

v2:
* Rename sock_path to uds_path.
* Update documentation to reflect when CAP_BPF is needed.
* Fix testpmd arguments in the provided example for Pods.
* Use AF_XDP API to update the xskmap entry.
---

Maryam Tahhan (3):
docs: AF_XDP Device Plugin
net/af_xdp: fix multi interface support for K8s net/af_xdp: support AF_XDP DP pinned maps

 doc/guides/howto/af_xdp_cni.rst        | 253 ------------------
 doc/guides/howto/af_xdp_dp.rst         | 340 +++++++++++++++++++++++++
 doc/guides/howto/index.rst             |   2 +-
 doc/guides/nics/af_xdp.rst             |  44 +++-
 doc/guides/rel_notes/release_24_03.rst |  17 ++
 drivers/net/af_xdp/rte_eth_af_xdp.c    | 167 ++++++++----
 6 files changed, 522 insertions(+), 301 deletions(-)
 delete mode 100644 doc/guides/howto/af_xdp_cni.rst
 create mode 100644 doc/guides/howto/af_xdp_dp.rst
Signature · 1 line
2.41.0
#1 lore raw

PATCH v11 1/3 docs: AF_XDP Device Plugin

Fixup the references to the AF_XDP Device Plugin in the documentation (was referred to as CNI previously) and document the single netdev limitation for deploying an AF_XDP based DPDK pod. Also renames af_xdp_cni.rst to af_xdp_dp.rst

Fixes 7fc6ae50369d ("net/af_xdp: support CNI Integration")
---
 doc/guides/howto/af_xdp_cni.rst | 253 ---------------------------
 doc/guides/howto/af_xdp_dp.rst  | 299 ++++++++++++++++++++++++++++++++
 doc/guides/howto/index.rst      |   2 +-
 doc/guides/nics/af_xdp.rst      |   4 +-
 4 files changed, 302 insertions(+), 256 deletions(-)
 delete mode 100644 doc/guides/howto/af_xdp_cni.rst
 create mode 100644 doc/guides/howto/af_xdp_dp.rst
doc/guides/howto/af_xdp_cni.rst deleted +0 -253
deleted file mode 100644
index a1a6d5b99c..0000000000
@@ -1,253 +0,0 @@
-.. SPDX-License-Identifier: BSD-3-Clause
-   Copyright(c) 2023 Intel Corporation.
-
-Using a CNI with the AF_XDP driver
-==================================
-
-Introduction
-------------
-
-CNI, the Container Network Interface, is a technology for configuring
-container network interfaces
-and which can be used to setup Kubernetes networking.
-AF_XDP is a Linux socket Address Family that enables an XDP program
-to redirect packets to a memory buffer in userspace.
-
-This document explains how to enable the `AF_XDP Plugin for Kubernetes`_ within
-a DPDK application using the :doc:`../nics/af_xdp` to connect and use these technologies.
-
-.. _AF_XDP Plugin for Kubernetes: https://github.com/intel/afxdp-plugins-for-kubernetes
-
-
-Background
-----------
-
-The standard :doc:`../nics/af_xdp` initialization process involves loading an eBPF program
-onto the kernel netdev to be used by the PMD.
-This operation requires root or escalated Linux privileges
-and thus prevents the PMD from working in an unprivileged container.
-The AF_XDP CNI plugin handles this situation
-by providing a device plugin that performs the program loading.
-
-At a technical level the CNI opens a Unix Domain Socket and listens for a client
-to make requests over that socket.
-A DPDK application acting as a client connects and initiates a configuration "handshake".
-The client then receives a file descriptor which points to the XSKMAP
-associated with the loaded eBPF program.
-The XSKMAP is a BPF map of AF_XDP sockets (XSK).
-The client can then proceed with creating an AF_XDP socket
-and inserting that socket into the XSKMAP pointed to by the descriptor.
-
-The EAL vdev argument ``use_cni`` is used to indicate that the user wishes
-to run the PMD in unprivileged mode and to receive the XSKMAP file descriptor
-from the CNI.
-When this flag is set,
-the ``XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD`` libbpf flag
-should be used when creating the socket
-to instruct libbpf not to load the default libbpf program on the netdev.
-Instead the loading is handled by the CNI.
-
-.. note::
-
-   The Unix Domain Socket file path appear in the end user is "/tmp/afxdp.sock".
-
-
-Prerequisites
--------------
-
-Docker and container prerequisites:
-
-* Set up the device plugin
-  as described in the instructions for `AF_XDP Plugin for Kubernetes`_.
-
-* The Docker image should contain the libbpf and libxdp libraries,
-  which are dependencies for AF_XDP,
-  and should include support for the ``ethtool`` command.
-
-* The Pod should have enabled the capabilities ``CAP_NET_RAW`` and ``CAP_BPF``
-  for AF_XDP along with support for hugepages.
-
-* Increase locked memory limit so containers have enough memory for packet buffers.
-  For example:
-
-  .. code-block:: console
-
-     cat << EOF | sudo tee /etc/systemd/system/containerd.service.d/limits.conf
-     [Service]
-     LimitMEMLOCK=infinity
-     EOF
-
-* dpdk-testpmd application should have AF_XDP feature enabled.
-
-  For further information see the docs for the: :doc:`../../nics/af_xdp`.
-
-
-Example
--------
-
-Howto run dpdk-testpmd with CNI plugin:
-
-* Clone the CNI plugin
-
-  .. code-block:: console
-
-     # git clone https://github.com/intel/afxdp-plugins-for-kubernetes.git
-
-* Build the CNI plugin
-
-  .. code-block:: console
-
-     # cd afxdp-plugins-for-kubernetes/
-     # make build
-
-  .. note::
-
-     CNI plugin has a dependence on the config.json.
-
-  Sample Config.json
-
-  .. code-block:: json
-
-     {
-        "logLevel":"debug",
-        "logFile":"afxdp-dp-e2e.log",
-        "pools":[
-           {
-              "name":"e2e",
-              "mode":"primary",
-              "timeout":30,
-              "ethtoolCmds" : ["-L -device- combined 1"],
-              "devices":[
-                 {
-                    "name":"ens785f0"
-                 }
-              ]
-           }
-        ]
-     }
-
-  For further reference please use the `config.json`_
-
-  .. _config.json: https://github.com/intel/afxdp-plugins-for-kubernetes/blob/v0.0.2/test/e2e/config.json
-
-* Create the Network Attachment definition
-
-  .. code-block:: console
-
-     # kubectl create -f nad.yaml
-
-  Sample nad.yml
-
-  .. code-block:: yaml
-
-      apiVersion: "k8s.cni.cncf.io/v1"
-      kind: NetworkAttachmentDefinition
-      metadata:
-        name: afxdp-e2e-test
-        annotations:
-          k8s.v1.cni.cncf.io/resourceName: afxdp/e2e
-      spec:
-        config: '{
-            "cniVersion": "0.3.0",
-            "type": "afxdp",
-            "mode": "cdq",
-            "logFile": "afxdp-cni-e2e.log",
-            "logLevel": "debug",
-            "ipam": {
-              "type": "host-local",
-              "subnet": "192.168.1.0/24",
-              "rangeStart": "192.168.1.200",
-              "rangeEnd": "192.168.1.216",
-              "routes": [
-                { "dst": "0.0.0.0/0" }
-              ],
-              "gateway": "192.168.1.1"
-            }
-          }'
-
-  For further reference please use the `nad.yaml`_
-
-  .. _nad.yaml: https://github.com/intel/afxdp-plugins-for-kubernetes/blob/v0.0.2/test/e2e/nad.yaml
-
-* Build the Docker image
-
-  .. code-block:: console
-
-     # docker build -t afxdp-e2e-test -f Dockerfile .
-
-  Sample Dockerfile:
-
-  .. code-block:: console
-
-     FROM ubuntu:20.04
-     RUN apt-get update -y
-     RUN apt install build-essential libelf-dev -y
-     RUN apt-get install iproute2  acl -y
-     RUN apt install python3-pyelftools ethtool -y
-     RUN apt install libnuma-dev libjansson-dev libpcap-dev net-tools -y
-     RUN apt-get install clang llvm -y
-     COPY ./libbpf<version>.tar.gz /tmp
-     RUN cd /tmp && tar -xvmf libbpf<version>.tar.gz && cd libbpf/src && make install
-     COPY ./libxdp<version>.tar.gz /tmp
-     RUN cd /tmp && tar -xvmf libxdp<version>.tar.gz && cd libxdp && make install
-
-  .. note::
-
-     All the files that need to COPY-ed should be in the same directory as the Dockerfile
-
-* Run the Pod
-
-  .. code-block:: console
-
-     # kubectl create -f pod.yaml
-
-  Sample pod.yaml:
-
-  .. code-block:: yaml
-
-     apiVersion: v1
-     kind: Pod
-     metadata:
-       name: afxdp-e2e-test
-       annotations:
-         k8s.v1.cni.cncf.io/networks: afxdp-e2e-test
-     spec:
-       containers:
-       - name: afxdp
-         image: afxdp-e2e-test:latest
-         imagePullPolicy: Never
-         env:
-         - name: LD_LIBRARY_PATH
-           value: /usr/lib64/:/usr/local/lib/
-         command: ["tail", "-f", "/dev/null"]
-         securityContext:
-          capabilities:
-             add:
-               - CAP_NET_RAW
-               - CAP_BPF
-         resources:
-           requests:
-             hugepages-2Mi: 2Gi
-             memory: 2Gi
-             afxdp/e2e: '1'
-           limits:
-             hugepages-2Mi: 2Gi
-             memory: 2Gi
-             afxdp/e2e: '1'
-
-  For further reference please use the `pod.yaml`_
-
-  .. _pod.yaml: https://github.com/intel/afxdp-plugins-for-kubernetes/blob/v0.0.2/test/e2e/pod-1c1d.yaml
-
-* Run DPDK with a command like the following:
-
-  .. code-block:: console
-
-     kubectl exec -i <Pod name> --container <containers name> -- \
-           /<Path>/dpdk-testpmd -l 0,1 --no-pci \
-           --vdev=net_af_xdp0,use_cni=1,iface=<interface name> \
-           -- --no-mlockall --in-memory
-
-For further reference please use the `e2e`_ test case in `AF_XDP Plugin for Kubernetes`_
-
-  .. _e2e: https://github.com/intel/afxdp-plugins-for-kubernetes/tree/v0.0.2/test/e2e
doc/guides/howto/af_xdp_dp.rst new file +299 -0

299 lines not on this page · raw message

doc/guides/howto/index.rst +1 -1
index 71a3381c36..a7692e8a97 100644
@@ -8,7 +8,7 @@ HowTo Guides
     :maxdepth: 2
     :numbered:
 
-    af_xdp_cni
+    af_xdp_dp
     lm_bond_virtio_sriov
     lm_virtio_vhost_user
     flow_bifurcation
doc/guides/nics/af_xdp.rst +2 -2
index 1932525d4d..4dd9c73742 100644
@@ -155,9 +155,9 @@ use_cni
 ~~~~~~~
 
 The EAL vdev argument ``use_cni`` is used to indicate that the user wishes to
-enable the `AF_XDP Plugin for Kubernetes`_ within a DPDK application.
+enable the `AF_XDP Device Plugin for Kubernetes`_ with a DPDK application/pod.
 
-.. _AF_XDP Plugin for Kubernetes: https://github.com/intel/afxdp-plugins-for-kubernetes
+.. _AF_XDP Device Plugin for Kubernetes: https://github.com/intel/afxdp-plugins-for-kubernetes
 
 .. code-block:: console
 
Signature · 1 line
2.41.0
#2 lore raw

PATCH v11 2/3 net/af_xdp: fix multi interface support for K8s

In reply to Maryam Tahhan

The original 'use_cni' implementation, was added to enable support for the AF_XDP PMD in a K8s env without any escalated privileges.
However 'use_cni' used a hardcoded socket rather than a configurable one. If a DPDK pod is requesting multiple net devices and these devices are from different pools, then the AF_XDP PMD attempts to mount all the netdev UDSes in the pod as /tmp/afxdp.sock. Which means that at best only 1 netdev will handshake correctly with the AF_XDP DP. This patch addresses this by making the socket parameter configurable using a new vdev param called 'dp_path' alongside the original 'use_cni' param. If the 'dp_path' parameter is not set alongside the 'use_cni' parameter, then it's configured inside the AF_XDP PMD (transparently to the user). This change has been tested
with the AF_XDP DP PR 81[1], with both single and multiple interfaces.

[1] https://github.com/intel/afxdp-plugins-for-kubernetes/pull/81

Fixes 7fc6ae50369d ("net/af_xdp: support CNI Integration")
---
 doc/guides/howto/af_xdp_dp.rst         | 62 +++++++++++------
 doc/guides/nics/af_xdp.rst             | 14 ++++
 doc/guides/rel_notes/release_24_03.rst |  7 ++
 drivers/net/af_xdp/rte_eth_af_xdp.c    | 94 ++++++++++++++++----------
 4 files changed, 121 insertions(+), 56 deletions(-)
doc/guides/howto/af_xdp_dp.rst +41 -21
index 7166d904bd..4aa6b5499f 100644
@@ -52,29 +52,33 @@ should be used when creating the socket
 to instruct libbpf not to load the default libbpf program on the netdev.
 Instead the loading is handled by the AF_XDP Device Plugin.
 
-Limitations
------------
+The EAL vdev argument ``dp_path`` is used alongside the ``use_cni`` argument
+to explicitly tell the AF_XDP PMD where to find the UDS to interact with the
+AF_XDP Device Plugin. If this argument is not passed alongside the ``use_cni``
+argument then the AF_XDP PMD configures it internally.
 
-For DPDK versions <= v23.11 the Unix Domain Socket file path appears in
-the pod at "/tmp/afxdp.sock". The handshake implementation in the AF_XDP PMD
-is only compatible with the AF_XDP Device Plugin up to commit id `38317c2`_
-and the pod is limited to a single netdev.
+.. note::
+
+    DPDK AF_XDP PMD <= v23.11 will only work with the AF_XDP Device Plugin
+    <= commit id `38317c2`_.
 
 .. note::
 
-    DPDK AF_XDP PMD <= v23.11 will not work with the latest version of the
-    AF_XDP Device Plugin.
+    DPDK AF_XDP PMD > v23.11 will work with latest version of the
+    AF_XDP Device Plugin through a combination of the ``dp_path`` and/or
+    the ``use_cni`` parameter. In these versions of the PMD if a user doesn't
+    explicitly set the ``dp_path`` parameter when using ``use_cni`` then that
+    path is transparently configured in the AF_XDP PMD to the default
+    `AF_XDP Device Plugin for Kubernetes`_ mount point path. The path can
+    be overridden by explicitly setting the ``dp_path`` param.
 
-The issue is if a single pod requests different devices from different pools it
-results in multiple UDS servers serving the pod with the container using only a
-single mount point for their UDS as ``/tmp/afxdp.sock``. This means that at best one
-device might be able to complete the handshake. This has been fixed in the AF_XDP
-Device Plugin so that the mount point in the pods for the UDS appear at
-``/tmp/afxdp_dp/<netdev>/afxdp.sock``. Later versions of DPDK fix this hardcoded path
-in the PMD alongside the ``use_cni`` parameter.
+.. note::
 
-.. _38317c2: https://github.com/intel/afxdp-plugins-for-kubernetes/commit/38317c256b5c7dfb39e013a0f76010c2ded03669
+    DPDK AF_XDP PMD > v23.11 is backwards compatible with (older) versions
+    of the AF_XDP DP <= commit id `38317c2`_ by explicitly setting ``dp_path`` to
+    ``/tmp/afxdp.sock``.
 
+.. _38317c2: https://github.com/intel/afxdp-plugins-for-kubernetes/commit/38317c256b5c7dfb39e013a0f76010c2ded03669
 
 Prerequisites
 -------------
@@ -105,10 +109,10 @@ Device Plugin and DPDK container prerequisites:
 
   .. code-block:: console
 
-     cat << EOF | sudo tee /etc/systemd/system/containerd.service.d/limits.conf
-     [Service]
-     LimitMEMLOCK=infinity
-     EOF
+    cat << EOF | sudo tee /etc/systemd/system/containerd.service.d/limits.conf
+    [Service]
+    LimitMEMLOCK=infinity
+    EOF
 
 * dpdk-testpmd application should have AF_XDP feature enabled.
 
@@ -284,7 +288,7 @@ Run dpdk-testpmd with the AF_XDP Device Plugin + CNI
         emptyDir:
           medium: HugePages
 
-  For further reference please use the `pod.yaml`_
+  For further reference please see the `pod.yaml`_
 
   .. _pod.yaml: https://github.com/intel/afxdp-plugins-for-kubernetes/blob/main/examples/pod-spec.yaml
 
@@ -297,3 +301,19 @@ Run dpdk-testpmd with the AF_XDP Device Plugin + CNI
            --vdev=net_af_xdp0,use_cni=1,iface=<interface name> \
            --no-mlockall --in-memory \
            -- -i --a --nb-cores=2 --rxq=1 --txq=1 --forward-mode=macswap;
+
+  Or
+
+  .. code-block:: console
+
+     kubectl exec -i <Pod name> --container <containers name> -- \
+           /<Path>/dpdk-testpmd -l 0,1 --no-pci \
+           --vdev=net_af_xdp0,use_cni=1,iface=<interface name>,dp_path="/tmp/afxdp_dp/<interface name>/afxdp.sock" \
+           --no-mlockall --in-memory \
+           -- -i --a --nb-cores=2 --rxq=1 --txq=1 --forward-mode=macswap;
+
+.. note::
+
+    If the ``dp_path`` parameter isn't explicitly set (like the example above)
+    the AF_XDP PMD will set the parameter value to
+    ``/tmp/afxdp_dp/<<interface name>>/afxdp.sock``.
doc/guides/nics/af_xdp.rst +14 -0
index 4dd9c73742..7f8651beda 100644
@@ -171,6 +171,20 @@ enable the `AF_XDP Device Plugin for Kubernetes`_ with a DPDK application/pod.
    so enabling and disabling of the promiscuous mode through the DPDK application
    is also not supported.
 
+dp_path
+~~~~~~~
+
+The EAL vdev argument ``dp_path`` is used alongside the ``use_cni`` argument
+to explicitly tell the AF_XDP PMD where to find the UDS to interact with the
+`AF_XDP Device Plugin for Kubernetes`_. If this argument is not passed
+alongside the ``use_cni`` argument then the AF_XDP PMD configures it internally.
+
+.. _AF_XDP Device Plugin for Kubernetes: https://github.com/intel/afxdp-plugins-for-kubernetes
+
+.. code-block:: console
+
+   --vdev=net_af_xdp0,use_cni=1,dp_path="/tmp/afxdp_dp/<<interface name>>/afxdp.sock"
+
 Limitations
 -----------
 
doc/guides/rel_notes/release_24_03.rst +7 -0
index 879bb4944c..b2b1f2566f 100644
@@ -138,6 +138,13 @@ New Features
     to support TLS v1.2, TLS v1.3 and DTLS v1.2.
   * Added PMD API to allow raw submission of instructions to CPT.
 
+* **Enabled AF_XDP PMD multi interface (UDS) support with AF_XDP Device Plugin**.
+
+  The EAL vdev argument for the AF_XDP PMD ``use_cni`` previously limited
+  a pod to using only a single netdev/interface. The latest changes (adding
+  the ``dp_path`` parameter) remove this limitation and maintain backward
+  compatibility for any applications already using the ``use_cni`` vdev
+  argument with the AF_XDP Device Plugin.
 
 Removed Items
 -------------
drivers/net/af_xdp/rte_eth_af_xdp.c +59 -35

268 lines not on this page · raw message

Signature · 1 line
2.41.0
#3 lore raw

snip

Quoted 23 lines
drivers/net/af_xdp/rte_eth_af_xdp.c +4 -7
@@ -1695,17 +1699,16 @@ xsk_configure(struct pmd_internals *internals,
 struct pkt_rx_queue *rxq,
 	}
 
 	if (internals->use_cni) {
-		int err, fd, map_fd;
+		int err, map_fd;
 
-		/* get socket fd from CNI plugin */
-		map_fd = get_cni_fd(internals->if_name);
+		/* get socket fd from AF_XDP Device Plugin */
+		map_fd = uds_get_xskmap_fd(internals->if_name, internals-
 >dp_path);
 		if (map_fd < 0) {
-			AF_XDP_LOG(ERR, "Failed to receive CNI plugin fd\n");
+			AF_XDP_LOG(ERR, "Failed to receive xskmap fd from
 AF_XDP Device Plugin\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);

drivers/net/af_xdp/rte_eth_af_xdp.c xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq, · line 1711

Quoted 2 lines
drivers/net/af_xdp/rte_eth_af_xdp.c +2 -0
+
+		err = xsk_socket__update_xskmap(rxq->xsk, map_fd);

Hi Maryam,

I've reviewed the series again. I haven't tested the device-plugin specific functionality as I don't have that environment set up, but outside of that I am happy that the new functionality doesn't break anything else. The doc updates look good to me now, thank you for the fixes.

I have just spotted one issue and I apologise for only catching it now.
Patch 2 introduces a dependency on the xsk_socket__update_xskmap function which is available in:
libbpf >= v0.3.0 and <= v0.6.0
libxdp > v1.2.0

The af_xdp.rst guide states we are compatible with libbpf (on it's own) <= v0.6.0. So users using libbpf < v0.3.0 will get an undefined reference warning for the xsk_socket__update_xskmap function.

Is it possible to implement fallback functionality (or if that's not possible, bail out) if that function is not available? See how this is done for the xsk_socket__create_shared function in meson.build and compat.h.

Thanks,
Ciara

Quoted 143 lines
(quoted diff) +0 -0
 		if (err) {
 			AF_XDP_LOG(ERR, "Failed to insert unprivileged xsk in

map.\n");

drivers/net/af_xdp/rte_eth_af_xdp.c +13 -7
 			goto out_xsk;
@@ -1881,13 +1884,13 @@ 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. Promiscuous enable and disable functionality
- * is removed.
+/* AF_XDP Device Plugin option works in unprivileged
+ * container environments and ethernet device functionality
+ * will be reduced. So additional customised eth_dev_ops
+ * struct is needed for the Device Plugin. Promiscuous
+ * enable and disable functionality is removed.
  **/
-static const struct eth_dev_ops ops_cni = {
+static const struct eth_dev_ops ops_afxdp_dp = {
 	.dev_start = eth_dev_start,
 	.dev_stop = eth_dev_stop,
 	.dev_close = eth_dev_close,
@@ -2023,7 +2026,8 @@ 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 *use_cni)
+		 int *busy_budget, int *force_copy, int *use_cni,
+		 char *dp_path)
 {
 	int ret;
@@ -2069,6 +2073,11 @@ 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_DP_PATH_ARG,
+				 &parse_prog_arg, dp_path);
+	if (ret < 0)
+		goto free_kvlist;
+
 free_kvlist:
 	rte_kvargs_free(kvlist);
drivers/net/af_xdp/rte_eth_af_xdp.c +2 -1
 	return ret;
@@ -2108,7 +2117,7 @@ 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 use_cni)
+	       int use_cni, const char *dp_path)
 {
 	const char *name = rte_vdev_device_name(dev);
 	const unsigned int numa_node = dev->device.numa_node;
@@ -2138,6 +2147,7 @@ init_internals(struct rte_vdev_device *dev, const
 char *if_name,
 	internals->shared_umem = shared_umem;
 	internals->force_copy = force_copy;
 	internals->use_cni = use_cni;
+	strlcpy(internals->dp_path, dp_path, PATH_MAX);
 
 	if (xdp_get_channels_info(if_name, &internals->max_queue_cnt,
drivers/net/af_xdp/rte_eth_af_xdp.c +1 -1
 				  &internals->combined_queue_cnt)) {
@@ -2199,7 +2209,7 @@ init_internals(struct rte_vdev_device *dev, const
 char *if_name,
 	if (!internals->use_cni)
 		eth_dev->dev_ops = &ops;
 	else
-		eth_dev->dev_ops = &ops_cni;
+		eth_dev->dev_ops = &ops_afxdp_dp;
 
 	eth_dev->rx_pkt_burst = eth_af_xdp_rx;
drivers/net/af_xdp/rte_eth_af_xdp.c +2 -1
 	eth_dev->tx_pkt_burst = eth_af_xdp_tx;
@@ -2328,6 +2338,7 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device
 *dev)
 	int busy_budget = -1, ret;
 	int force_copy = 0;
 	int use_cni = 0;
+	char dp_path[PATH_MAX] = {'\0'};
 	struct rte_eth_dev *eth_dev = NULL;
 	const char *name = rte_vdev_device_name(dev);
@@ -2370,7 +2381,7 @@ 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, &use_cni) < 0) {
+			     &busy_budget, &force_copy, &use_cni, dp_path) <
 0) {
 		AF_XDP_LOG(ERR, "Invalid kvargs value\n");
drivers/net/af_xdp/rte_eth_af_xdp.c +5 -1
 		return -EINVAL;
 	}
@@ -2384,7 +2395,19 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device
 *dev)
 	if (use_cni && strnlen(prog_path, PATH_MAX)) {
 		AF_XDP_LOG(ERR, "When '%s' parameter is used, '%s'
 parameter is not valid\n",
 			ETH_AF_XDP_USE_CNI_ARG,
 ETH_AF_XDP_PROG_ARG);
-			return -EINVAL;
+		return -EINVAL;
+	}
+
+	if (use_cni && !strnlen(dp_path, PATH_MAX)) {
+		snprintf(dp_path, sizeof(dp_path), "%s/%s/%s",

DP_BASE_PATH, if_name, DP_UDS_SOCK);

(quoted diff) +1 -0
+		AF_XDP_LOG(INFO, "'%s' parameter not provided, setting

value to '%s'\n",

drivers/net/af_xdp/rte_eth_af_xdp.c +5 -0
+			ETH_AF_XDP_DP_PATH_ARG, dp_path);
+	}
+
+	if (!use_cni && strnlen(dp_path, PATH_MAX)) {
+		AF_XDP_LOG(ERR, "'%s' parameter is set, but '%s' was not

enabled\n",

drivers/net/af_xdp/rte_eth_af_xdp.c +1 -0
+			ETH_AF_XDP_DP_PATH_ARG,

ETH_AF_XDP_USE_CNI_ARG);

drivers/net/af_xdp/rte_eth_af_xdp.c +2 -1
+		return -EINVAL;
 	}
 
 	if (strlen(if_name) == 0) {
@@ -2410,7 +2433,7 @@ rte_pmd_af_xdp_probe(struct rte_vdev_device
 *dev)
 
 	eth_dev = init_internals(dev, if_name, xsk_start_queue_idx,
 				 xsk_queue_cnt, shared_umem, prog_path,
-				 busy_budget, force_copy, use_cni);
+				 busy_budget, force_copy, use_cni, dp_path);
 	if (eth_dev == NULL) {
 		AF_XDP_LOG(ERR, "Failed to init internals\n");
drivers/net/af_xdp/rte_eth_af_xdp.c +2 -2
 		return -1;
@@ -2471,4 +2494,5 @@ RTE_PMD_REGISTER_PARAM_STRING(net_af_xdp,
 			      "xdp_prog=<string> "
 			      "busy_budget=<int> "
 			      "force_copy=<int> "
-			      "use_cni=<int> ");
+			      "use_cni=<int> "
+			      "dp_path=<string> ");
--

2.41.0

#4 lore raw
Quoted 36 lines · Ciara

snip

Quoted 23 lines
drivers/net/af_xdp/rte_eth_af_xdp.c +4 -7
@@ -1695,17 +1699,16 @@ xsk_configure(struct pmd_internals *internals,
 struct pkt_rx_queue *rxq,
  	}
 
  	if (internals->use_cni) {
-		int err, fd, map_fd;
+		int err, map_fd;
 
-		/* get socket fd from CNI plugin */
-		map_fd = get_cni_fd(internals->if_name);
+		/* get socket fd from AF_XDP Device Plugin */
+		map_fd = uds_get_xskmap_fd(internals->if_name, internals-
 > dp_path);
  		if (map_fd < 0) {
-			AF_XDP_LOG(ERR, "Failed to receive CNI plugin fd\n");
+			AF_XDP_LOG(ERR, "Failed to receive xskmap fd from
 AF_XDP Device Plugin\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);

drivers/net/af_xdp/rte_eth_af_xdp.c xsk_configure(struct pmd_internals *internals, struct pkt_rx_queue *rxq, · line 1711

Quoted 2 lines
drivers/net/af_xdp/rte_eth_af_xdp.c +2 -0
+
+		err = xsk_socket__update_xskmap(rxq->xsk, map_fd);

Hi Maryam,

I've reviewed the series again. I haven't tested the device-plugin specific functionality as I don't have that environment set up, but outside of that I am happy that the new functionality doesn't break anything else. The doc updates look good to me now, thank you for the fixes.

I have just spotted one issue and I apologise for only catching it now.
Patch 2 introduces a dependency on the xsk_socket__update_xskmap function which is available in:
libbpf >= v0.3.0 and <= v0.6.0
libxdp > v1.2.0

Quoted 7 lines

The af_xdp.rst guide states we are compatible with libbpf (on it's own) <= v0.6.0. So users using libbpf < v0.3.0 will get an undefined reference warning for the xsk_socket__update_xskmap function.

Is it possible to implement fallback functionality (or if that's not possible, bail out) if that function is not available? See how this is done for the xsk_socket__create_shared function in meson.build and compat.h.

Thanks,
Ciara

Hi Ciara

Yeah, no prob - I can re introduce the ``bpf_map_update_elem()`` call as a fall back. I will need to retest the permissions as I remember escalated permissions being required when issuing that call directly in the PMD which is why I moved it to using ``xsk_socket__update_xskmap()`` but let me retest and circle back with some changes.

:) Looks like we are going to a v12 :)

BR Maryam

#5 lore raw

PATCH v11 3/3 net/af_xdp: support AF_XDP DP pinned maps

In reply to Maryam Tahhan

Enable the AF_XDP PMD to retrieve the xskmap from a pinned eBPF map. This map is expected to be pinned by an external entity like the AF_XDP Device Plugin. This enabled unprivileged pods to create and use AF_XDP sockets.

---
 doc/guides/howto/af_xdp_dp.rst         | 35 ++++++++--
 doc/guides/nics/af_xdp.rst             | 34 ++++++++--
 doc/guides/rel_notes/release_24_03.rst | 10 +++
 drivers/net/af_xdp/rte_eth_af_xdp.c    | 93 ++++++++++++++++++++------
 4 files changed, 141 insertions(+), 31 deletions(-)
doc/guides/howto/af_xdp_dp.rst +28 -7
index 4aa6b5499f..8b9b5ebbad 100644
@@ -52,10 +52,21 @@ should be used when creating the socket
 to instruct libbpf not to load the default libbpf program on the netdev.
 Instead the loading is handled by the AF_XDP Device Plugin.
 
-The EAL vdev argument ``dp_path`` is used alongside the ``use_cni`` argument
-to explicitly tell the AF_XDP PMD where to find the UDS to interact with the
-AF_XDP Device Plugin. If this argument is not passed alongside the ``use_cni``
-argument then the AF_XDP PMD configures it internally.
+The EAL vdev argument ``use_pinned_map`` is used indicate to the AF_XDP PMD to
+retrieve the XSKMAP fd from a pinned eBPF map. This map is expected to be pinned
+by an external entity like the AF_XDP Device Plugin. This enabled unprivileged pods
+to create and use AF_XDP sockets. When this flag is set, the
+``XSK_LIBBPF_FLAGS__INHIBIT_PROG_LOAD`` libbpf flag is used by the AF_XDP PMD when
+creating the AF_XDP socket.
+
+The EAL vdev argument ``dp_path`` is used alongside the ``use_cni`` or ``use_pinned_map``
+arguments to explicitly tell the AF_XDP PMD where to find either:
+
+1. The UDS to interact with the AF_XDP Device Plugin. OR
+2. The pinned xskmap to use when creating AF_XDP sockets.
+
+If this argument is not passed alongside the ``use_cni`` or ``use_pinned_map`` arguments then
+the AF_XDP PMD configures it internally to the `AF_XDP Device Plugin for Kubernetes`_.
 
 .. note::
 
@@ -312,8 +323,18 @@ Run dpdk-testpmd with the AF_XDP Device Plugin + CNI
            --no-mlockall --in-memory \
            -- -i --a --nb-cores=2 --rxq=1 --txq=1 --forward-mode=macswap;
 
+  Or
+
+  .. code-block:: console
+
+     kubectl exec -i <Pod name> --container <containers name> -- \
+           /<Path>/dpdk-testpmd -l 0,1 --no-pci \
+           --vdev=net_af_xdp0,use_pinned_map=1,iface=<interface name>,dp_path="/tmp/afxdp_dp/<interface name>/xsks_map" \
+           --no-mlockall --in-memory \
+           -- -i --a --nb-cores=2 --rxq=1 --txq=1 --forward-mode=macswap;
+
 .. note::
 
-    If the ``dp_path`` parameter isn't explicitly set (like the example above)
-    the AF_XDP PMD will set the parameter value to
-    ``/tmp/afxdp_dp/<<interface name>>/afxdp.sock``.
+    If the ``dp_path`` parameter isn't explicitly set with ``use_cni`` or ``use_pinned_map``
+    the AF_XDP PMD will set the parameter values to the `AF_XDP Device Plugin for Kubernetes`_
+    defaults.
doc/guides/nics/af_xdp.rst +30 -4
index 7f8651beda..940bbf60f2 100644
@@ -171,13 +171,35 @@ enable the `AF_XDP Device Plugin for Kubernetes`_ with a DPDK application/pod.
    so enabling and disabling of the promiscuous mode through the DPDK application
    is also not supported.
 
+use_pinned_map
+~~~~~~~~~~~~~~
+
+The EAL vdev argument ``use_pinned_map`` is used to indicate that the user wishes to
+load a pinned xskmap mounted by `AF_XDP Device Plugin for Kubernetes`_ in the DPDK
+application/pod.
+
+.. _AF_XDP Device Plugin for Kubernetes: https://github.com/intel/afxdp-plugins-for-kubernetes
+
+.. code-block:: console
+
+   --vdev=net_af_xdp0,use_pinned_map=1
+
+.. note::
+
+    This feature can also be used with any external entity that can pin an eBPF map, not just
+    the `AF_XDP Device Plugin for Kubernetes`_.
+
 dp_path
 ~~~~~~~
 
-The EAL vdev argument ``dp_path`` is used alongside the ``use_cni`` argument
-to explicitly tell the AF_XDP PMD where to find the UDS to interact with the
-`AF_XDP Device Plugin for Kubernetes`_. If this argument is not passed
-alongside the ``use_cni`` argument then the AF_XDP PMD configures it internally.
+The EAL vdev argument ``dp_path`` is used alongside the ``use_cni`` or ``use_pinned_map``
+arguments to explicitly tell the AF_XDP PMD where to find either:
+
+1. The UDS to interact with the AF_XDP Device Plugin. OR
+2. The pinned xskmap to use when creating AF_XDP sockets.
+
+If this argument is not passed alongside the ``use_cni`` or ``use_pinned_map`` arguments then
+the AF_XDP PMD configures it internally to the `AF_XDP Device Plugin for Kubernetes`_.
 
 .. _AF_XDP Device Plugin for Kubernetes: https://github.com/intel/afxdp-plugins-for-kubernetes
 
@@ -185,6 +207,10 @@ alongside the ``use_cni`` argument then the AF_XDP PMD configures it internally.
 
    --vdev=net_af_xdp0,use_cni=1,dp_path="/tmp/afxdp_dp/<<interface name>>/afxdp.sock"
 
+.. code-block:: console
+
+   --vdev=net_af_xdp0,use_pinned_map=1,dp_path="/tmp/afxdp_dp/<<interface name>>/xsks_map"
+
 Limitations
 -----------
 
doc/guides/rel_notes/release_24_03.rst +10 -0
index b2b1f2566f..95d9a0f842 100644
@@ -146,6 +146,16 @@ New Features
   compatibility for any applications already using the ``use_cni`` vdev
   argument with the AF_XDP Device Plugin.
 
+* **Integrated AF_XDP PMD with AF_XDP Device Plugin eBPF map pinning support**.
+
+  The EAL vdev argument for the AF_XDP PMD ``use_map_pinning`` was added
+  to allow Kubernetes Pods to use AF_XDP with DPDK, and run  with limited
+  privileges, without having to do a full handshake over a Unix Domain
+  Socket with the Device Plugin. This flag indicates that the AF_XDP PMD
+  will be used in unprivileged mode and will obtain the XSKMAP FD by calling
+  ``bpf_obj_get()`` for an xskmap pinned (by the AF_XDP DP) inside the
+  container.
+
 Removed Items
 -------------
 
drivers/net/af_xdp/rte_eth_af_xdp.c +73 -20

203 lines not on this page · raw message

Signature · 1 line
2.41.0