introduce fail-safe PMD

Introduce the fail-safe poll mode driver initialization and enable its build infrastructure.

This PMD allows for applications to benefit from true hot-plugging support without having to implement it.

It intercepts and manages Ethernet device removal events issued by slave PMDs and re-initializes them transparently when brought back. It also allows defining a contingency to the removal of a device, by designating a fail-over device that will take on transmitting operations if the preferred device is removed.

Applications only see a fail-safe instance, without caring for underlying activity ensuring their continued operations.

Signed-off-by Gaetan Rivet <[email protected]>

Acked-by Olga Shern <[email protected]>

---
 MAINTAINERS                                       |   5 +
 config/common_base                                |   6 +
 doc/guides/nics/fail_safe.rst                     | 133 +++++
 doc/guides/nics/features/failsafe.ini             |  24 +
 doc/guides/nics/index.rst                         |   1 +
 drivers/net/Makefile                              |   2 +
 drivers/net/failsafe/Makefile                     |  76 +++
 drivers/net/failsafe/failsafe.c                   | 231 ++++++++
 drivers/net/failsafe/failsafe_args.c              | 331 +++++++++++
 drivers/net/failsafe/failsafe_eal.c               | 154 +++++
 drivers/net/failsafe/failsafe_ops.c               | 663 ++++++++++++++++++++++
 drivers/net/failsafe/failsafe_private.h           | 227 ++++++++
 drivers/net/failsafe/failsafe_rxtx.c              | 107 ++++
 drivers/net/failsafe/rte_pmd_failsafe_version.map |   4 +
 mk/rte.app.mk                                     |   1 +
 15 files changed, 1965 insertions(+)
 create mode 100644 doc/guides/nics/fail_safe.rst
 create mode 100644 doc/guides/nics/features/failsafe.ini
 create mode 100644 drivers/net/failsafe/Makefile
 create mode 100644 drivers/net/failsafe/failsafe.c
 create mode 100644 drivers/net/failsafe/failsafe_args.c
 create mode 100644 drivers/net/failsafe/failsafe_eal.c
 create mode 100644 drivers/net/failsafe/failsafe_ops.c
 create mode 100644 drivers/net/failsafe/failsafe_private.h
 create mode 100644 drivers/net/failsafe/failsafe_rxtx.c
 create mode 100644 drivers/net/failsafe/rte_pmd_failsafe_version.map

MAINTAINERS +5 -0

index 8fb2132..b4a446f 100644
@@ -336,6 +336,11 @@ F: drivers/net/enic/
 F: doc/guides/nics/enic.rst
 F: doc/guides/nics/features/enic.ini
 
+Fail-safe PMD
+M: Gaetan Rivet <[email protected]>
+F: drivers/net/failsafe/
+F: doc/guides/nics/fail_safe.rst
+
 Intel e1000
 M: Wenzhuo Lu <[email protected]>
 F: drivers/net/e1000/

config/common_base +6 -0

index bb1ba8b..cf5e7f5 100644
@@ -420,6 +420,12 @@ CONFIG_RTE_LIBRTE_PMD_XENVIRT=n
 CONFIG_RTE_LIBRTE_PMD_NULL=y
 
 #
+# Compile fail-safe PMD
+#
+CONFIG_RTE_LIBRTE_PMD_FAILSAFE=y
+CONFIG_RTE_LIBRTE_PMD_FAILSAFE_DEBUG=n
+
+#
 # Do prefetch of packet data within PMD driver receive function
 #
 CONFIG_RTE_PMD_PACKET_PREFETCH=y

doc/guides/nics/fail_safe.rst new file +133 -0

new file mode 100644
index 0000000..056f85f
@@ -0,0 +1,133 @@
+..  BSD LICENSE
+    Copyright 2017 6WIND S.A.
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions
+    are met:
+
+    * Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above copyright
+    notice, this list of conditions and the following disclaimer in
+    the documentation and/or other materials provided with the
+    distribution.
+    * Neither the name of 6WIND S.A. nor the names of its
+    contributors may be used to endorse or promote products derived
+    from this software without specific prior written permission.
+
+    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+    OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+    SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+    LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+Fail-safe poll mode driver library
+==================================
+
+The Fail-safe poll mode driver library (**librte_pmd_failsafe**) is a virtual
+device that allows using any device supporting hotplug (sudden device removal
+and plugging on its bus), without modifying other components relying on such
+device (application, other PMDs).
+
+Additionally to the Seamless Hotplug feature, the Fail-safe PMD offers the
+ability to redirect operations to secondary devices when the primary has been
+removed from the system.
+
+.. note::
+
+   The library is enabled by default. You can enable it or disable it manually
+   by setting the ``CONFIG_RTE_LIBRTE_PMD_FAILSAFE`` configuration option.
+
+Features
+--------
+
+The Fail-safe PMD only supports a limited set of features. If you plan to use a
+device underneath the Fail-safe PMD with a specific feature, this feature must
+be supported by the Fail-safe PMD to avoid throwing any error.
+
+Check the feature matrix for the complete set of supported features.
+
+Compilation options
+-------------------
+
+These options can be modified in the ``$RTE_TARGET/build/.config`` file.
+
+- ``CONFIG_RTE_LIBRTE_PMD_FAILSAFE`` (default **y**)
+
+  Toggle compiling librte_pmd_failsafe itself.
+
+- ``CONFIG_RTE_LIBRTE_PMD_FAILSAFE_DEBUG`` (default **n**)
+
+  Toggle debugging code.
+
+Using the Fail-safe PMD from the EAL command line
+-------------------------------------------------
+
+The Fail-safe PMD can be used like most other DPDK virtual devices, by passing a
+``--vdev`` parameter to the EAL when starting the application. The device name
+must start with the *net_failsafe* prefix, followed by numbers or letters. This
+name must be unique for each device. Each fail-safe instance must have at least one
+sub-device, up to ``RTE_MAX_ETHPORTS-1``.
+
+A sub-device can be any legal DPDK device, including possibly another fail-safe
+instance.
+
+Fail-safe command line parameters
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+- **dev(<iface>)** parameter
+
+  This parameter allows the user to define a sub-device. The ``<iface>`` part of
+  this parameter must be a valid device definition. It could be the argument
+  provided to a ``-w`` PCI device specification or the argument that would be
+  given to a ``--vdev`` parameter (including a fail-safe).
+  Enclosing the device definition within parenthesis here allows using
+  additional sub-device parameters if need be. They will be passed on to the
+  sub-device.
+
+- **mac** parameter [MAC address]
+
+  This parameter allows the user to set a default MAC address to the fail-safe
+  and all of its sub-devices.
+  If no default mac address is provided, the fail-safe PMD will read the MAC
+  address of the first of its sub-device to be successfully probed and use it as
+  its default MAC address, trying to set it to all of its other sub-devices.
+  If no sub-device was successfully probed at initialization, then a random MAC
+  address is generated, that will be subsequently applied to all sub-device once
+  they are probed.
+
+Usage example
+~~~~~~~~~~~~~
+
+This section shows some example of using **testpmd** with a fail-safe PMD.
+
+#. Request huge pages:
+
+   .. code-block:: console
+
+      echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages
+
+#. Start testpmd
+
+   .. code-block:: console
+
+      $RTE_TARGET/build/app/testpmd -c 0xff -n 4 --no-pci \
+         --vdev='net_failsafe0,mac=de:ad:be:ef:01:02,dev(84:00.0),dev(net_ring0,nodeaction=r1:0:CREATE)' -- \
+         -i
+
+Using the Fail-safe PMD from an application
+-------------------------------------------
+
+This driver strives to be as seamless as possible to existing applications, in
+order to propose the hotplug functionality in the easiest way possible.
+
+Care must be taken, however, to respect the **ether** API concerning device
+access, and in particular, using the ``RTE_ETH_FOREACH_DEV`` macro to iterate
+over ethernet devices, instead of directly accessing them or by writing one's
+own device iterator.

doc/guides/nics/features/failsafe.ini new file +24 -0

new file mode 100644
index 0000000..3c52823
@@ -0,0 +1,24 @@
+;
+; Supported features of the 'fail-safe' poll mode driver.
+;
+; Refer to default.ini for the full list of available PMD features.
+;
+[Features]
+Link status          = Y
+Queue start/stop     = Y
+MTU update           = Y
+Jumbo frame          = Y
+Promiscuous mode     = Y
+Allmulticast mode    = Y
+Unicast MAC filter   = Y
+Multicast MAC filter = Y
+VLAN filter          = Y
+Packet type parsing  = Y
+Basic stats          = Y
+Stats per queue      = Y
+ARMv7                = Y
+ARMv8                = Y
+Power8               = Y
+x86-32               = Y
+x86-64               = Y
+Usage doc            = Y

doc/guides/nics/index.rst +1 -0

index 240d082..17eaaf4 100644
@@ -64,6 +64,7 @@ Network Interface Controller Drivers
     vhost
     vmxnet3
     pcap_ring
+    fail_safe
 
 **Figures**
 

drivers/net/Makefile +2 -0

index 35ed813..d33c959 100644
@@ -59,6 +59,8 @@ DIRS-$(CONFIG_RTE_LIBRTE_ENA_PMD) += ena
 DEPDIRS-ena = $(core-libs)
 DIRS-$(CONFIG_RTE_LIBRTE_ENIC_PMD) += enic
 DEPDIRS-enic = $(core-libs) librte_hash
+DIRS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe
+DEPDIRS-failsafe = $(core-libs)
 DIRS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k
 DEPDIRS-fm10k = $(core-libs) librte_hash
 DIRS-$(CONFIG_RTE_LIBRTE_I40E_PMD) += i40e

drivers/net/failsafe/Makefile new file +76 -0

new file mode 100644
index 0000000..c759035
@@ -0,0 +1,76 @@
+#   BSD LICENSE
+#
+#   Copyright 2017 6WIND S.A.
+#   Copyright 2017 Mellanox.
+#
+#   Redistribution and use in source and binary forms, with or without
+#   modification, are permitted provided that the following conditions
+#   are met:
+#
+#     * Redistributions of source code must retain the above copyright
+#       notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in
+#       the documentation and/or other materials provided with the
+#       distribution.
+#     * Neither the name of 6WIND S.A. nor the names of its
+#       contributors may be used to endorse or promote products derived
+#       from this software without specific prior written permission.
+#
+#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+# Library name
+LIB = librte_pmd_failsafe.a
+
+EXPORT_MAP := rte_pmd_failsafe_version.map
+
+LIBABIVER := 1
+
+# Sources are stored in SRCS-y
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_args.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_eal.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_ops.c
+SRCS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += failsafe_rxtx.c
+
+# No exported include files
+
+# This lib depends upon:
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += lib/librte_eal
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += lib/librte_ether
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += lib/librte_kvargs
+DEPDIRS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE) += lib/librte_mbuf
+
+ifneq ($(DEBUG),)
+CONFIG_RTE_LIBRTE_PMD_FAILSAFE_DEBUG := y
+endif
+
+# Basic CFLAGS:
+CFLAGS += -std=gnu99 -Wall -Wextra
+CFLAGS += -I.
+CFLAGS += -D_DEFAULT_SOURCE
+CFLAGS += -D_XOPEN_SOURCE=700
+CFLAGS += $(WERROR_FLAGS)
+CFLAGS += -Wno-strict-prototypes
+CFLAGS += -pedantic -DPEDANTIC
+
+ifeq ($(CONFIG_RTE_LIBRTE_PMD_FAILSAFE_DEBUG),y)
+CFLAGS += -g -UNDEBUG
+else
+CFLAGS += -O3
+CFLAGS += -DNDEBUG
+endif
+
+include $(RTE_SDK)/mk/rte.lib.mk

drivers/net/failsafe/failsafe.c new file +231 -0

new file mode 100644
index 0000000..7cf33e8
@@ -0,0 +1,231 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright 2017 6WIND S.A.
+ *   Copyright 2017 Mellanox.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of 6WIND S.A. nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#include <rte_alarm.h>
+#include <rte_malloc.h>
+#include <rte_ethdev.h>
+#include <rte_ethdev_vdev.h>
+#include <rte_devargs.h>
+#include <rte_kvargs.h>
+#include <rte_vdev.h>
+
+#include "failsafe_private.h"
+
+const char pmd_failsafe_driver_name[] = FAILSAFE_DRIVER_NAME;
+static const struct rte_eth_link eth_link = {
+       .link_speed = ETH_SPEED_NUM_10G,
+       .link_duplex = ETH_LINK_FULL_DUPLEX,
+       .link_status = ETH_LINK_UP,
+       .link_autoneg = ETH_LINK_SPEED_AUTONEG,
+};
+
+static int
+fs_sub_device_create(struct rte_eth_dev *dev,
+               const char *params)
+{
+       uint8_t nb_subs;
+       int ret;
+
+       ret = failsafe_args_count_subdevice(dev, params);
+       if (ret)
+               return ret;
+       if (PRIV(dev)->subs_tail > FAILSAFE_MAX_ETHPORTS) {
+               ERROR("Cannot allocate more than %d ports",
+                       FAILSAFE_MAX_ETHPORTS);
+               return -ENOSPC;
+       }
+       nb_subs = PRIV(dev)->subs_tail;
+       PRIV(dev)->subs = rte_zmalloc(NULL,
+                       sizeof(struct sub_device) * nb_subs,
+                       RTE_CACHE_LINE_SIZE);
+       if (PRIV(dev)->subs == NULL) {
+               ERROR("Could not allocate sub_devices");
+               return -ENOMEM;
+       }
+       return 0;
+}
+
+static void
+fs_sub_device_free(struct rte_eth_dev *dev)
+{
+       rte_free(PRIV(dev)->subs);
+}
+
+static int
+fs_eth_dev_create(struct rte_vdev_device *vdev)
+{
+       struct rte_eth_dev *dev;
+       struct ether_addr *mac;
+       struct fs_priv *priv;
+       struct sub_device *sdev;
+       const char *params;
+       unsigned int socket_id;
+       uint8_t i;
+       int ret;
+
+       dev = NULL;
+       priv = NULL;
+       params = rte_vdev_device_args(vdev);
+       socket_id = rte_socket_id();
+       INFO("Creating fail-safe device on NUMA socket %u",
+            socket_id);
+       dev = rte_eth_vdev_allocate(vdev, sizeof(*priv));
+       if (dev == NULL) {
+               ERROR("Unable to allocate rte_eth_dev");
+               return -1;
+       }
+       priv = dev->data->dev_private;
+       PRIV(dev)->dev = dev;
+       dev->dev_ops = &failsafe_ops;
+       TAILQ_INIT(&dev->link_intr_cbs);
+       dev->data->dev_flags = 0x0;
+       dev->data->mac_addrs = &PRIV(dev)->mac_addrs[0];
+       dev->data->dev_link = eth_link;
+       PRIV(dev)->nb_mac_addr = 1;
+       dev->rx_pkt_burst = (eth_rx_burst_t)&failsafe_rx_burst;
+       dev->tx_pkt_burst = (eth_tx_burst_t)&failsafe_tx_burst;
+       if (params == NULL) {
+               ERROR("This PMD requires sub-devices, none provided");
+               goto free_dev;
+       }
+       ret = fs_sub_device_create(dev, params);
+       if (ret) {
+               ERROR("Could not allocate sub_devices");
+               goto free_dev;
+       }
+       ret = failsafe_args_parse(dev, params);
+       if (ret)
+               goto free_subs;
+       ret = failsafe_eal_init(dev);
+       if (ret)
+               goto free_args;
+       mac = &dev->data->mac_addrs[0];
+       if (mac_from_arg) {
+               /*
+                * If MAC address was provided as a parameter,
+                * apply to all probed slaves.
+                */
+               FOREACH_SUBDEV_ST(sdev, i, dev, DEV_PROBED) {
+                       ret = rte_eth_dev_default_mac_addr_set(PORT_ID(sdev),
+                                                              mac);
+                       if (ret) {
+                               ERROR("Failed to set default MAC address");
+                               goto free_args;
+                       }
+               }
+       } else {
+               /*
+                * Use the ether_addr from first probed
+                * device, either preferred or fallback.
+                */
+               FOREACH_SUBDEV(sdev, i, dev)
+                       if (sdev->state >= DEV_PROBED) {
+                               ether_addr_copy(&ETH(sdev)->data->mac_addrs[0],
+                                               mac);
+                               break;
+                       }
+               /*
+                * If no device has been probed and no ether_addr
+                * has been provided on the command line, use a random
+                * valid one.
+                * It will be applied during future slave state syncs to
+                * probed slaves.
+                */
+               if (i == priv->subs_tail)
+                       eth_random_addr(&mac->addr_bytes[0]);
+       }
+       INFO("MAC address is %02x:%02x:%02x:%02x:%02x:%02x",
+               mac->addr_bytes[0], mac->addr_bytes[1],
+               mac->addr_bytes[2], mac->addr_bytes[3],
+               mac->addr_bytes[4], mac->addr_bytes[5]);
+       return 0;
+free_args:
+       failsafe_args_free(dev);
+free_subs:
+       fs_sub_device_free(dev);
+free_dev:
+       rte_eth_dev_release_port(dev);
+       return -1;
+}
+
+static int
+fs_rte_eth_free(const char *name)
+{
+       struct rte_eth_dev *dev;
+       int ret;
+
+       dev = rte_eth_dev_allocated(name);
+       if (dev == NULL)
+               return -ENODEV;
+       ret = failsafe_eal_uninit(dev);
+       if (ret)
+               ERROR("Error while uninitializing sub-EAL");
+       failsafe_args_free(dev);
+       fs_sub_device_free(dev);
+       rte_free(PRIV(dev));
+       rte_eth_dev_release_port(dev);
+       return ret;
+}
+
+static int
+rte_pmd_failsafe_probe(struct rte_vdev_device *vdev)
+{
+       const char *name;
+
+       name = rte_vdev_device_name(vdev);
+       if (vdev == NULL)
+               return -EINVAL;
+       INFO("Initializing " FAILSAFE_DRIVER_NAME " for %s",
+                       name);
+       return fs_eth_dev_create(vdev);
+}
+
+static int
+rte_pmd_failsafe_remove(struct rte_vdev_device *vdev)
+{
+       const char *name;
+
+       name = rte_vdev_device_name(vdev);
+       if (name == NULL)
+               return -EINVAL;
+       INFO("Uninitializing " FAILSAFE_DRIVER_NAME " for %s", name);
+       return fs_rte_eth_free(name);
+}
+
+static struct rte_vdev_driver failsafe_drv = {
+       .probe = rte_pmd_failsafe_probe,
+       .remove = rte_pmd_failsafe_remove,
+};
+
+RTE_PMD_REGISTER_VDEV(net_failsafe, failsafe_drv);
+RTE_PMD_REGISTER_ALIAS(net_failsafe, eth_failsafe);
+RTE_PMD_REGISTER_PARAM_STRING(net_failsafe, PMD_FAILSAFE_PARAM_STRING);

drivers/net/failsafe/failsafe_args.c new file +83 -0

new file mode 100644
index 0000000..f07d26e
@@ -0,0 +1,331 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright 2017 6WIND S.A.
+ *   Copyright 2017 Mellanox.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of 6WIND S.A. nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#include <string.h>
+#include <errno.h>
+
+#include <rte_devargs.h>
+#include <rte_malloc.h>
+#include <rte_kvargs.h>
+
+#include "failsafe_private.h"
+
+#define DEVARGS_MAXLEN 4096
+
+/* Callback used when a new device is found in devargs */
+typedef int (parse_cb)(struct rte_eth_dev *dev, const char *params,
+               uint8_t head);
+
+int mac_from_arg;
+
+const char *pmd_failsafe_init_parameters[] = {
+       PMD_FAILSAFE_MAC_KVARG,
+       NULL,
+};
+
+/*
+ * input: text.
+ * output: 0: if text[0] != '(',
+ *         0: if there are no corresponding ')'
+ *         n: distance to corresponding ')' otherwise
+ */
+static size_t
+closing_paren(const char *text)
+{
+       int nb_open = 0;
+       size_t i = 0;
+
+       while (text[i] != '\0') {
+               if (text[i] == '(')
+                       nb_open++;
+               if (text[i] == ')')
+                       nb_open--;
+               if (nb_open == 0)
+                       return i;
+               i++;
+       }
+       return 0;
+}
+
+static int
+fs_parse_device(struct sub_device *sdev, char *args)
+{
+       struct rte_devargs *d;
+       int ret;

drivers/net/failsafe/failsafe_args.c new file +238 -0

+       return 0;
+}
+
+static int
+fs_parse_device_param(struct rte_eth_dev *dev, const char *param,
+               uint8_t head)
+{
+       struct fs_priv *priv;
+       struct sub_device *sdev;
+       char *args = NULL;
+       size_t a, b;
+       int ret;
+
+       priv = PRIV(dev);
+       a = 0;
+       b = 0;
+       ret = 0;
+       while  (param[b] != '(' &&
+               param[b] != '\0')
+               b++;
+       a = b;
+       b += closing_paren(&param[b]);
+       if (a == b) {
+               ERROR("Dangling parenthesis");
+               return -EINVAL;
+       }
+       a += 1;
+       args = strndup(&param[a], b - a);
+       if (args == NULL) {
+               ERROR("Not enough memory for parameter parsing");
+               return -ENOMEM;
+       }
+       sdev = &priv->subs[head];
+       if (strncmp(param, "dev", 3) == 0) {
+               ret = fs_parse_device(sdev, args);
+               if (ret)
+                       goto free_args;
+       } else {
+               ERROR("Unrecognized device type: %.*s", (int)b, param);
+               return -EINVAL;
+       }
+free_args:
+       free(args);
+       return ret;
+}
+
+static int
+fs_parse_sub_devices(parse_cb *cb,
+               struct rte_eth_dev *dev, const char *params)
+{
+       size_t a, b;
+       uint8_t head;
+       int ret;
+
+       a = 0;
+       head = 0;
+       ret = 0;
+       while (params[a] != '\0') {
+               b = a;
+               while (params[b] != '(' &&
+                      params[b] != ',' &&
+                      params[b] != '\0')
+                       b++;
+               if (b == a) {
+                       ERROR("Invalid parameter");
+                       return -EINVAL;
+               }
+               if (params[b] == ',') {
+                       a = b + 1;
+                       continue;
+               }
+               if (params[b] == '(') {
+                       size_t start = b;
+
+                       b += closing_paren(&params[b]);
+                       if (b == start) {
+                               ERROR("Dangling parenthesis");
+                               return -EINVAL;
+                       }
+                       ret = (*cb)(dev, &params[a], head);
+                       if (ret)
+                               return ret;
+                       head += 1;
+                       b += 1;
+                       if (params[b] == '\0')
+                               return 0;
+               }
+               a = b + 1;
+       }
+       return 0;
+}
+
+static int
+fs_remove_sub_devices_definition(char params[DEVARGS_MAXLEN])
+{
+       char buffer[DEVARGS_MAXLEN] = {0};
+       size_t a, b;
+       int i;
+
+       a = 0;
+       i = 0;
+       while (params[a] != '\0') {
+               b = a;
+               while (params[b] != '(' &&
+                      params[b] != ',' &&
+                      params[b] != '\0')
+                       b++;
+               if (b == a) {
+                       ERROR("Invalid parameter");
+                       return -EINVAL;
+               }
+               if (params[b] == ',' || params[b] == '\0')
+                       i += snprintf(&buffer[i], b - a + 1, "%s", &params[a]);
+               if (params[b] == '(') {
+                       size_t start = b;
+                       b += closing_paren(&params[b]);
+                       if (b == start)
+                               return -EINVAL;
+                       b += 1;
+                       if (params[b] == '\0')
+                               goto out;
+               }
+               a = b + 1;
+       }
+out:
+       snprintf(params, DEVARGS_MAXLEN, "%s", buffer);
+       return 0;
+}
+
+static int
+fs_get_mac_addr_arg(const char *key __rte_unused,
+               const char *value, void *out)
+{
+       struct ether_addr *ea = out;
+       int ret;
+
+       if ((value == NULL) || (out == NULL))
+               return -EINVAL;
+       ret = sscanf(value, "%hhx:%hhx:%hhx:%hhx:%hhx:%hhx",
+               &ea->addr_bytes[0], &ea->addr_bytes[1],
+               &ea->addr_bytes[2], &ea->addr_bytes[3],
+               &ea->addr_bytes[4], &ea->addr_bytes[5]);
+       return ret != ETHER_ADDR_LEN;
+}
+
+int
+failsafe_args_parse(struct rte_eth_dev *dev, const char *params)
+{
+       struct fs_priv *priv;
+       char mut_params[DEVARGS_MAXLEN] = "";
+       struct rte_kvargs *kvlist = NULL;
+       unsigned int arg_count;
+       size_t n;
+       int ret;
+
+       if (dev == NULL || params == NULL)
+               return -EINVAL;
+       priv = PRIV(dev);
+       ret = 0;
+       priv->subs_tx = FAILSAFE_MAX_ETHPORTS;
+       /* default parameters */
+       mac_from_arg = 0;
+       n = snprintf(mut_params, sizeof(mut_params), "%s", params);
+       if (n >= sizeof(mut_params)) {
+               ERROR("Parameter string too long (>=%zu)",
+                               sizeof(mut_params));
+               return -ENOMEM;
+       }
+       ret = fs_parse_sub_devices(fs_parse_device_param,
+                                  dev, params);
+       if (ret < 0)
+               return ret;
+       ret = fs_remove_sub_devices_definition(mut_params);
+       if (ret < 0)
+               return ret;
+       if (strnlen(mut_params, sizeof(mut_params)) > 0) {
+               kvlist = rte_kvargs_parse(mut_params,
+                               pmd_failsafe_init_parameters);
+               if (kvlist == NULL) {
+                       ERROR("Error parsing parameters, usage:\n"
+                               PMD_FAILSAFE_PARAM_STRING);
+                       return -1;
+               }
+               /* MAC addr */
+               arg_count = rte_kvargs_count(kvlist,
+                               PMD_FAILSAFE_MAC_KVARG);
+               if (arg_count == 1) {
+                       ret = rte_kvargs_process(kvlist,
+                                       PMD_FAILSAFE_MAC_KVARG,
+                                       &fs_get_mac_addr_arg,
+                                       &dev->data->mac_addrs[0]);
+                       if (ret < 0)
+                               goto free_kvlist;
+                       mac_from_arg = 1;
+               }
+       }
+free_kvlist:
+       rte_kvargs_free(kvlist);
+       return ret;
+}
+
+void
+failsafe_args_free(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       FOREACH_SUBDEV(sdev, i, dev) {
+               free(sdev->devargs.args);
+               sdev->devargs.args = NULL;
+       }
+}
+
+static int
+fs_count_device(struct rte_eth_dev *dev, const char *param,
+               uint8_t head __rte_unused)
+{
+       size_t b = 0;
+
+       while  (param[b] != '(' &&
+               param[b] != '\0')
+               b++;
+       if (strncmp(param, "dev", b) &&
+           strncmp(param, "exec", b)) {
+               ERROR("Unrecognized device type: %.*s", (int)b, param);
+               return -EINVAL;
+       }
+       PRIV(dev)->subs_tail += 1;
+       return 0;
+}
+
+int
+failsafe_args_count_subdevice(struct rte_eth_dev *dev,
+                       const char *params)
+{
+       return fs_parse_sub_devices(fs_count_device,
+                                   dev, params);
+}

drivers/net/failsafe/failsafe_eal.c new file +40 -0

new file mode 100644
index 0000000..6c3a811
@@ -0,0 +1,154 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright 2017 6WIND S.A.
+ *   Copyright 2017 Mellanox.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of 6WIND S.A. nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <rte_malloc.h>
+
+#include "failsafe_private.h"
+
+static struct rte_eth_dev *
+fs_find_ethdev(const struct rte_device *dev)
+{

drivers/net/failsafe/failsafe_eal.c new file +7 -0

+
+static int
+fs_bus_init(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       struct rte_device *rdev;
+       struct rte_devargs *da;

drivers/net/failsafe/failsafe_eal.c new file +87 -0

+               ret = rdev ? 0 : -rte_errno;
+               if (ret) {
+                       ERROR("sub_device %d probe failed %s%s%s", i,
+                             errno ? "(" : "",
+                             errno ? strerror(rte_errno) : "",
+                             errno ? ")" : "");
+                       continue;
+               }
+               ETH(sdev) = fs_find_ethdev(rdev);
+               if (ETH(sdev) == NULL) {
+                       ERROR("sub_device %d init went wrong", i);
+                       return -ENODEV;
+               }
+               sdev->dev = ETH(sdev)->device;
+               ETH(sdev)->state = RTE_ETH_DEV_DEFERRED;
+               sdev->state = DEV_PROBED;
+       }
+       return 0;
+}
+
+int
+failsafe_eal_init(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int ret;
+
+       ret = fs_bus_init(dev);
+       if (ret)
+               return ret;
+       /*
+        * We only update TX_SUBDEV if we are not started.
+        * If a sub_device is emitting, we will switch the TX_SUBDEV to the
+        * preferred port only upon starting it, so that the switch is smoother.
+        */
+       if (PREFERRED_SUBDEV(dev)->state >= DEV_PROBED) {
+               if (TX_SUBDEV(dev) != PREFERRED_SUBDEV(dev) &&
+                   (TX_SUBDEV(dev) == NULL ||
+                    (TX_SUBDEV(dev) && TX_SUBDEV(dev)->state < DEV_STARTED))) {
+                       DEBUG("Switching tx_dev to preferred sub_device");
+                       PRIV(dev)->subs_tx = 0;
+               }
+       } else {
+               if ((TX_SUBDEV(dev) && TX_SUBDEV(dev)->state < DEV_PROBED) ||
+                   TX_SUBDEV(dev) == NULL) {
+                       /* Using first probed device */
+                       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_PROBED) {
+                               DEBUG("Switching tx_dev to sub_device %d",
+                                     i);
+                               PRIV(dev)->subs_tx = i;
+                               break;
+                       }
+               }
+       }
+       return 0;
+}
+
+static int
+fs_bus_uninit(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev = NULL;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_PROBED) {
+               ret = rte_eal_hotplug_remove(sdev->bus->name,
+                                            sdev->dev->name);
+               if (ret) {
+                       ERROR("Failed to remove requested device %s",
+                             sdev->dev->name);
+                       continue;
+               }
+               sdev->state = DEV_PROBED - 1;
+       }
+       return 0;
+}
+
+int
+failsafe_eal_uninit(struct rte_eth_dev *dev)
+{
+       int ret;
+
+       ret = fs_bus_uninit(dev);
+       if (ret)
+               return ret;
+       return 0;
+}

drivers/net/failsafe/failsafe_ops.c new file +663 -0

new file mode 100644
index 0000000..693162e
@@ -0,0 +1,663 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright 2017 6WIND S.A.
+ *   Copyright 2017 Mellanox.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of 6WIND S.A. nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <assert.h>
+#include <stdint.h>
+#include <rte_ethdev.h>
+#include <rte_malloc.h>
+
+#include "failsafe_private.h"
+
+static struct rte_eth_dev_info default_infos = {
+       .driver_name = pmd_failsafe_driver_name,
+       /* Max possible number of elements */
+       .max_rx_pktlen = UINT32_MAX,
+       .max_rx_queues = RTE_MAX_QUEUES_PER_PORT,
+       .max_tx_queues = RTE_MAX_QUEUES_PER_PORT,
+       .max_mac_addrs = FAILSAFE_MAX_ETHADDR,
+       .max_hash_mac_addrs = UINT32_MAX,
+       .max_vfs = UINT16_MAX,
+       .max_vmdq_pools = UINT16_MAX,
+       .rx_desc_lim = {
+               .nb_max = UINT16_MAX,
+               .nb_min = 0,
+               .nb_align = 1,
+               .nb_seg_max = UINT16_MAX,
+               .nb_mtu_seg_max = UINT16_MAX,
+       },
+       .tx_desc_lim = {
+               .nb_max = UINT16_MAX,
+               .nb_min = 0,
+               .nb_align = 1,
+               .nb_seg_max = UINT16_MAX,
+               .nb_mtu_seg_max = UINT16_MAX,
+       },
+       /* Set of understood capabilities */
+       .rx_offload_capa = 0x0,
+       .tx_offload_capa = 0x0,
+       .flow_type_rss_offloads = 0x0,
+};
+
+static int
+fs_dev_configure(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV(sdev, i, dev) {
+               if (sdev->state != DEV_PROBED)
+                       continue;
+               DEBUG("Configuring sub-device %d", i);
+               ret = rte_eth_dev_configure(PORT_ID(sdev),
+                                       dev->data->nb_rx_queues,
+                                       dev->data->nb_tx_queues,
+                                       &dev->data->dev_conf);
+               if (ret) {
+                       ERROR("Could not configure sub_device %d", i);
+                       return ret;
+               }
+               sdev->state = DEV_ACTIVE;
+       }
+       return 0;
+}
+
+static int
+fs_dev_start(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV(sdev, i, dev) {
+               if (sdev->state != DEV_ACTIVE)
+                       continue;
+               DEBUG("Starting sub_device %d", i);
+               ret = rte_eth_dev_start(PORT_ID(sdev));
+               if (ret)
+                       return ret;
+               sdev->state = DEV_STARTED;
+       }
+       if (PREFERRED_SUBDEV(dev)->state == DEV_STARTED) {
+               if (TX_SUBDEV(dev) != PREFERRED_SUBDEV(dev)) {
+                       DEBUG("Switching tx_dev to preferred sub_device");
+                       PRIV(dev)->subs_tx = 0;
+               }
+       } else {
+               if ((TX_SUBDEV(dev) && TX_SUBDEV(dev)->state < DEV_STARTED) ||
+                   TX_SUBDEV(dev) == NULL) {
+                       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_STARTED) {
+                               DEBUG("Switching tx_dev to sub_device %d", i);
+                               PRIV(dev)->subs_tx = i;
+                               break;
+                       }
+               }
+       }
+       return 0;
+}
+
+static void
+fs_dev_stop(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_STARTED) {
+               rte_eth_dev_stop(PORT_ID(sdev));
+               sdev->state = DEV_STARTED - 1;
+       }
+}
+
+static int
+fs_dev_set_link_up(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               DEBUG("Calling rte_eth_dev_set_link_up on sub_device %d", i);
+               ret = rte_eth_dev_set_link_up(PORT_ID(sdev));
+               if (ret) {
+                       ERROR("Operation rte_eth_dev_set_link_up failed for sub_device %d"
+                             " with error %d", i, ret);
+                       return ret;
+               }
+       }
+       return 0;
+}
+
+static int
+fs_dev_set_link_down(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               DEBUG("Calling rte_eth_dev_set_link_down on sub_device %d", i);
+               ret = rte_eth_dev_set_link_down(PORT_ID(sdev));
+               if (ret) {
+                       ERROR("Operation rte_eth_dev_set_link_down failed for sub_device %d"
+                             " with error %d", i, ret);
+                       return ret;
+               }
+       }
+       return 0;
+}
+
+static void fs_dev_free_queues(struct rte_eth_dev *dev);
+static void
+fs_dev_close(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               DEBUG("Closing sub_device %d", i);
+               rte_eth_dev_close(PORT_ID(sdev));
+               sdev->state = DEV_ACTIVE - 1;
+       }
+       fs_dev_free_queues(dev);
+}
+
+static void
+fs_rx_queue_release(void *queue)
+{
+       struct rte_eth_dev *dev;
+       struct sub_device *sdev;
+       uint8_t i;
+       struct rxq *rxq;
+
+       if (queue == NULL)
+               return;
+       rxq = queue;
+       dev = rxq->priv->dev;
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE)
+               SUBOPS(sdev, rx_queue_release)
+                       (ETH(sdev)->data->rx_queues[rxq->qid]);
+       dev->data->rx_queues[rxq->qid] = NULL;
+       rte_free(rxq);
+}
+
+static int
+fs_rx_queue_setup(struct rte_eth_dev *dev,
+               uint16_t rx_queue_id,
+               uint16_t nb_rx_desc,
+               unsigned int socket_id,
+               const struct rte_eth_rxconf *rx_conf,
+               struct rte_mempool *mb_pool)
+{
+       struct sub_device *sdev;
+       struct rxq *rxq;
+       uint8_t i;
+       int ret;
+
+       rxq = dev->data->rx_queues[rx_queue_id];
+       if (rxq != NULL) {
+               fs_rx_queue_release(rxq);
+               dev->data->rx_queues[rx_queue_id] = NULL;
+       }
+       rxq = rte_zmalloc(NULL, sizeof(*rxq),
+                         RTE_CACHE_LINE_SIZE);
+       if (rxq == NULL)
+               return -ENOMEM;
+       rxq->qid = rx_queue_id;
+       rxq->socket_id = socket_id;
+       rxq->info.mp = mb_pool;
+       rxq->info.conf = *rx_conf;
+       rxq->info.nb_desc = nb_rx_desc;
+       rxq->priv = PRIV(dev);
+       dev->data->rx_queues[rx_queue_id] = rxq;
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               ret = rte_eth_rx_queue_setup(PORT_ID(sdev),
+                               rx_queue_id,
+                               nb_rx_desc, socket_id,
+                               rx_conf, mb_pool);
+               if (ret) {
+                       ERROR("RX queue setup failed for sub_device %d", i);
+                       goto free_rxq;
+               }
+       }
+       return 0;
+free_rxq:
+       fs_rx_queue_release(rxq);
+       return ret;
+}
+
+static void
+fs_tx_queue_release(void *queue)
+{
+       struct rte_eth_dev *dev;
+       struct sub_device *sdev;
+       uint8_t i;
+       struct txq *txq;
+
+       if (queue == NULL)
+               return;
+       txq = queue;
+       dev = txq->priv->dev;
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE)
+               SUBOPS(sdev, tx_queue_release)
+                       (ETH(sdev)->data->tx_queues[txq->qid]);
+       dev->data->tx_queues[txq->qid] = NULL;
+       rte_free(txq);
+}
+
+static int
+fs_tx_queue_setup(struct rte_eth_dev *dev,
+               uint16_t tx_queue_id,
+               uint16_t nb_tx_desc,
+               unsigned int socket_id,
+               const struct rte_eth_txconf *tx_conf)
+{
+       struct sub_device *sdev;
+       struct txq *txq;
+       uint8_t i;
+       int ret;
+
+       txq = dev->data->tx_queues[tx_queue_id];
+       if (txq != NULL) {
+               fs_tx_queue_release(txq);
+               dev->data->tx_queues[tx_queue_id] = NULL;
+       }
+       txq = rte_zmalloc("ethdev TX queue", sizeof(*txq),
+                         RTE_CACHE_LINE_SIZE);
+       if (txq == NULL)
+               return -ENOMEM;
+       txq->qid = tx_queue_id;
+       txq->socket_id = socket_id;
+       txq->info.conf = *tx_conf;
+       txq->info.nb_desc = nb_tx_desc;
+       txq->priv = PRIV(dev);
+       dev->data->tx_queues[tx_queue_id] = txq;
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               ret = rte_eth_tx_queue_setup(PORT_ID(sdev),
+                               tx_queue_id,
+                               nb_tx_desc, socket_id,
+                               tx_conf);
+               if (ret) {
+                       ERROR("TX queue setup failed for sub_device %d", i);
+                       goto free_txq;
+               }
+       }
+       return 0;
+free_txq:
+       fs_tx_queue_release(txq);
+       return ret;
+}
+
+static void
+fs_dev_free_queues(struct rte_eth_dev *dev)
+{
+       uint16_t i;
+
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               fs_rx_queue_release(dev->data->rx_queues[i]);
+               dev->data->rx_queues[i] = NULL;
+       }
+       dev->data->nb_rx_queues = 0;
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               fs_tx_queue_release(dev->data->tx_queues[i]);
+               dev->data->tx_queues[i] = NULL;
+       }
+       dev->data->nb_tx_queues = 0;
+}
+
+static void
+fs_promiscuous_enable(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE)
+               rte_eth_promiscuous_enable(PORT_ID(sdev));
+}
+
+static void
+fs_promiscuous_disable(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE)
+               rte_eth_promiscuous_disable(PORT_ID(sdev));
+}
+
+static void
+fs_allmulticast_enable(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE)
+               rte_eth_allmulticast_enable(PORT_ID(sdev));
+}
+
+static void
+fs_allmulticast_disable(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE)
+               rte_eth_allmulticast_disable(PORT_ID(sdev));
+}
+
+static int
+fs_link_update(struct rte_eth_dev *dev,
+               int wait_to_complete)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               DEBUG("Calling link_update on sub_device %d", i);
+               ret = (SUBOPS(sdev, link_update))(ETH(sdev), wait_to_complete);
+               if (ret && ret != -1) {
+                       ERROR("Link update failed for sub_device %d with error %d",
+                             i, ret);
+                       return ret;
+               }
+       }
+       if (TX_SUBDEV(dev)) {
+               struct rte_eth_link *l1;
+               struct rte_eth_link *l2;
+
+               l1 = &dev->data->dev_link;
+               l2 = &ETH(TX_SUBDEV(dev))->data->dev_link;
+               if (memcmp(l1, l2, sizeof(*l1))) {
+                       *l1 = *l2;
+                       return 0;
+               }
+       }
+       return -1;
+}
+
+static void
+fs_stats_get(struct rte_eth_dev *dev,
+            struct rte_eth_stats *stats)
+{
+       memset(stats, 0, sizeof(*stats));
+       if (TX_SUBDEV(dev) == NULL)
+               return;
+       rte_eth_stats_get(PORT_ID(TX_SUBDEV(dev)), stats);
+}
+
+static void
+fs_stats_reset(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE)
+               rte_eth_stats_reset(PORT_ID(sdev));
+}
+
+/**
+ * Fail-safe dev_infos_get rules:
+ *
+ * No sub_device:
+ *   Numerables:
+ *      Use the maximum possible values for any field, so as not
+ *      to impede any further configuration effort.
+ *   Capabilities:
+ *      Limits capabilities to those that are understood by the
+ *      fail-safe PMD. This understanding stems from the fail-safe
+ *      being capable of verifying that the related capability is
+ *      expressed within the device configuration (struct rte_eth_conf).
+ *
+ * At least one probed sub_device:
+ *   Numerables:
+ *      Uses values from the active probed sub_device
+ *      The rationale here is that if any sub_device is less capable
+ *      (for example concerning the number of queues) than the active
+ *      sub_device, then its subsequent configuration will fail.
+ *      It is impossible to foresee this failure when the failing sub_device
+ *      is supposed to be plugged-in later on, so the configuration process
+ *      is the single point of failure and error reporting.
+ *   Capabilities:
+ *      Uses a logical AND of RX capabilities among
+ *      all sub_devices and the default capabilities.
+ *      Uses a logical AND of TX capabilities among
+ *      the active probed sub_device and the default capabilities.
+ *
+ */
+static void
+fs_dev_infos_get(struct rte_eth_dev *dev,
+                 struct rte_eth_dev_info *infos)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       sdev = TX_SUBDEV(dev);
+       if (sdev == NULL) {
+               DEBUG("No probed device, using default infos");
+               rte_memcpy(&PRIV(dev)->infos, &default_infos,
+                          sizeof(default_infos));
+       } else {
+               uint32_t rx_offload_capa;
+
+               rx_offload_capa = default_infos.rx_offload_capa;
+               FOREACH_SUBDEV_ST(sdev, i, dev, DEV_PROBED) {
+                       rte_eth_dev_info_get(PORT_ID(sdev),
+                                       &PRIV(dev)->infos);
+                       rx_offload_capa &= PRIV(dev)->infos.rx_offload_capa;
+               }
+               sdev = TX_SUBDEV(dev);
+               rte_eth_dev_info_get(PORT_ID(sdev), &PRIV(dev)->infos);
+               PRIV(dev)->infos.rx_offload_capa = rx_offload_capa;
+               PRIV(dev)->infos.tx_offload_capa &=
+                                       default_infos.tx_offload_capa;
+               PRIV(dev)->infos.flow_type_rss_offloads &=
+                                       default_infos.flow_type_rss_offloads;
+       }
+       rte_memcpy(infos, &PRIV(dev)->infos, sizeof(*infos));
+}
+
+static const uint32_t *
+fs_dev_supported_ptypes_get(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       struct rte_eth_dev *edev;
+
+       sdev = TX_SUBDEV(dev);
+       if (sdev == NULL)
+               return NULL;
+       edev = ETH(sdev);
+       /* ENOTSUP: counts as no supported ptypes */
+       if (SUBOPS(sdev, dev_supported_ptypes_get) == NULL)
+               return NULL;
+       /*
+        * The API does not permit to do a clean AND of all ptypes,
+        * It is also incomplete by design and we do not really care
+        * to have a best possible value in this context.
+        * We just return the ptypes of the device of highest
+        * priority, usually the PREFERRED device.
+        */
+       return SUBOPS(sdev, dev_supported_ptypes_get)(edev);
+}
+
+static int
+fs_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               DEBUG("Calling rte_eth_dev_set_mtu on sub_device %d", i);
+               ret = rte_eth_dev_set_mtu(PORT_ID(sdev), mtu);
+               if (ret) {
+                       ERROR("Operation rte_eth_dev_set_mtu failed for sub_device %d"
+                             " with error %d", i, ret);
+                       return ret;
+               }
+       }
+       return 0;
+}
+
+static int
+fs_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               DEBUG("Calling rte_eth_dev_vlan_filter on sub_device %d", i);
+               ret = rte_eth_dev_vlan_filter(PORT_ID(sdev), vlan_id, on);
+               if (ret) {
+                       ERROR("Operation rte_eth_dev_vlan_filter failed for sub_device %d"
+                             " with error %d", i, ret);
+                       return ret;
+               }
+       }
+       return 0;
+}
+
+static int
+fs_flow_ctrl_get(struct rte_eth_dev *dev,
+               struct rte_eth_fc_conf *fc_conf)
+{
+       struct sub_device *sdev;
+
+       sdev = TX_SUBDEV(dev);
+       if (sdev == NULL)
+               return 0;
+       if (SUBOPS(sdev, flow_ctrl_get) == NULL)
+               return -ENOTSUP;
+       return SUBOPS(sdev, flow_ctrl_get)(ETH(sdev), fc_conf);
+}
+
+static int
+fs_flow_ctrl_set(struct rte_eth_dev *dev,
+               struct rte_eth_fc_conf *fc_conf)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               DEBUG("Calling rte_eth_dev_flow_ctrl_set on sub_device %d", i);
+               ret = rte_eth_dev_flow_ctrl_set(PORT_ID(sdev), fc_conf);
+               if (ret) {
+                       ERROR("Operation rte_eth_dev_flow_ctrl_set failed for sub_device %d"
+                             " with error %d", i, ret);
+                       return ret;
+               }
+       }
+       return 0;
+}
+
+static void
+fs_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       /* No check: already done within the rte_eth_dev_mac_addr_remove
+        * call for the fail-safe device.
+        */
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE)
+               rte_eth_dev_mac_addr_remove(PORT_ID(sdev),
+                               &dev->data->mac_addrs[index]);
+       PRIV(dev)->mac_addr_pool[index] = 0;
+}
+
+static int
+fs_mac_addr_add(struct rte_eth_dev *dev,
+               struct ether_addr *mac_addr,
+               uint32_t index,
+               uint32_t vmdq)
+{
+       struct sub_device *sdev;
+       int ret;
+       uint8_t i;
+
+       assert(index < FAILSAFE_MAX_ETHADDR);
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE) {
+               ret = rte_eth_dev_mac_addr_add(PORT_ID(sdev), mac_addr, vmdq);
+               if (ret) {
+                       ERROR("Operation rte_eth_dev_mac_addr_add failed for sub_device %"
+                             PRIu8 " with error %d", i, ret);
+                       return ret;
+               }
+       }
+       if (index >= PRIV(dev)->nb_mac_addr) {
+               DEBUG("Growing mac_addrs array");
+               PRIV(dev)->nb_mac_addr = index;
+       }
+       PRIV(dev)->mac_addr_pool[index] = vmdq;
+       return 0;
+}
+
+static void
+fs_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
+{
+       struct sub_device *sdev;
+       uint8_t i;
+
+       FOREACH_SUBDEV_ST(sdev, i, dev, DEV_ACTIVE)
+               rte_eth_dev_default_mac_addr_set(PORT_ID(sdev), mac_addr);
+}
+
+const struct eth_dev_ops failsafe_ops = {
+       .dev_configure = fs_dev_configure,
+       .dev_start = fs_dev_start,
+       .dev_stop = fs_dev_stop,
+       .dev_set_link_down = fs_dev_set_link_down,
+       .dev_set_link_up = fs_dev_set_link_up,
+       .dev_close = fs_dev_close,
+       .promiscuous_enable = fs_promiscuous_enable,
+       .promiscuous_disable = fs_promiscuous_disable,
+       .allmulticast_enable = fs_allmulticast_enable,
+       .allmulticast_disable = fs_allmulticast_disable,
+       .link_update = fs_link_update,
+       .stats_get = fs_stats_get,
+       .stats_reset = fs_stats_reset,
+       .dev_infos_get = fs_dev_infos_get,
+       .dev_supported_ptypes_get = fs_dev_supported_ptypes_get,
+       .mtu_set = fs_mtu_set,
+       .vlan_filter_set = fs_vlan_filter_set,
+       .rx_queue_setup = fs_rx_queue_setup,
+       .tx_queue_setup = fs_tx_queue_setup,
+       .rx_queue_release = fs_rx_queue_release,
+       .tx_queue_release = fs_tx_queue_release,
+       .flow_ctrl_get = fs_flow_ctrl_get,
+       .flow_ctrl_set = fs_flow_ctrl_set,
+       .mac_addr_remove = fs_mac_addr_remove,
+       .mac_addr_add = fs_mac_addr_add,
+       .mac_addr_set = fs_mac_addr_set,
+};

drivers/net/failsafe/failsafe_private.h new file +227 -0

new file mode 100644
index 0000000..e7a7592
@@ -0,0 +1,227 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright 2017 6WIND S.A.
+ *   Copyright 2017 Mellanox.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of 6WIND S.A. nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _RTE_ETH_FAILSAFE_PRIVATE_H_
+#define _RTE_ETH_FAILSAFE_PRIVATE_H_
+
+#include <rte_dev.h>
+#include <rte_ethdev.h>
+#include <rte_devargs.h>
+
+#define FAILSAFE_DRIVER_NAME "Fail-safe PMD"
+
+#define PMD_FAILSAFE_MAC_KVARG "mac"
+#define PMD_FAILSAFE_PARAM_STRING      \
+       "dev(<ifc>),"                   \
+       "mac=mac_addr"                  \
+       ""
+
+#define FAILSAFE_PLUGIN_DEFAULT_TIMEOUT_MS 2000
+
+#define FAILSAFE_MAX_ETHPORTS 2
+#define FAILSAFE_MAX_ETHADDR 128
+
+/* TYPES */
+
+struct rxq {
+       struct fs_priv *priv;
+       uint16_t qid;
+       /* id of last sub_device polled */
+       uint8_t last_polled;
+       unsigned int socket_id;
+       struct rte_eth_rxq_info info;
+};
+
+struct txq {
+       struct fs_priv *priv;
+       uint16_t qid;
+       unsigned int socket_id;
+       struct rte_eth_txq_info info;
+};
+
+enum dev_state {
+       DEV_UNDEFINED = 0,
+       DEV_PARSED,
+       DEV_PROBED,
+       DEV_ACTIVE,
+       DEV_STARTED,
+};
+
+struct sub_device {
+       /* Exhaustive DPDK device description */
+       struct rte_devargs devargs;
+       struct rte_bus *bus;
+       struct rte_device *dev;
+       struct rte_eth_dev *edev;
+       /* Device state machine */
+       enum dev_state state;
+};
+
+struct fs_priv {
+       struct rte_eth_dev *dev;
+       /*
+        * Set of sub_devices.
+        * subs[0] is the preferred device
+        * any other is just another slave
+        */
+       struct sub_device *subs;
+       uint8_t subs_head; /* if head == tail, no subs */
+       uint8_t subs_tail; /* first invalid */
+       uint8_t subs_tx; /* current emitting device */
+       uint8_t current_probed;
+       /* current number of mac_addr slots allocated. */
+       uint32_t nb_mac_addr;
+       struct ether_addr mac_addrs[FAILSAFE_MAX_ETHADDR];
+       uint32_t mac_addr_pool[FAILSAFE_MAX_ETHADDR];
+       /* current capabilities */
+       struct rte_eth_dev_info infos;
+};
+
+/* RX / TX */
+
+uint16_t failsafe_rx_burst(void *rxq,
+               struct rte_mbuf **rx_pkts, uint16_t nb_pkts);
+uint16_t failsafe_tx_burst(void *txq,
+               struct rte_mbuf **tx_pkts, uint16_t nb_pkts);
+
+/* ARGS */
+
+int failsafe_args_parse(struct rte_eth_dev *dev, const char *params);
+void failsafe_args_free(struct rte_eth_dev *dev);
+int failsafe_args_count_subdevice(struct rte_eth_dev *dev, const char *params);
+
+/* EAL */
+
+int failsafe_eal_init(struct rte_eth_dev *dev);
+int failsafe_eal_uninit(struct rte_eth_dev *dev);
+
+/* GLOBALS */
+
+extern const char pmd_failsafe_driver_name[];
+extern const struct eth_dev_ops failsafe_ops;
+extern int mac_from_arg;
+
+/* HELPERS */
+
+/* dev: (struct rte_eth_dev *) fail-safe device */
+#define PRIV(dev) \
+       ((struct fs_priv *)(dev)->data->dev_private)
+
+/* sdev: (struct sub_device *) */
+#define ETH(sdev) \
+       ((sdev)->edev)
+
+/* sdev: (struct sub_device *) */
+#define PORT_ID(sdev) \
+       (ETH(sdev)->data->port_id)
+
+/**
+ * Stateful iterator construct over fail-safe sub-devices:
+ * s:     (struct sub_device *), iterator
+ * i:     (uint8_t), increment
+ * dev:   (struct rte_eth_dev *), fail-safe ethdev
+ * state: (enum dev_state), minimum acceptable device state
+ */
+#define FOREACH_SUBDEV_ST(s, i, dev, state)                            \
+       for (i = fs_find_next((dev), 0, state);                         \
+            i < PRIV(dev)->subs_tail && (s = &PRIV(dev)->subs[i]);     \
+            i = fs_find_next((dev), i + 1, state))
+
+/**
+ * Iterator construct over fail-safe sub-devices:
+ * s:   (struct sub_device *), iterator
+ * i:   (uint8_t), increment
+ * dev: (struct rte_eth_dev *), fail-safe ethdev
+ */
+#define FOREACH_SUBDEV(s, i, dev)                      \
+       FOREACH_SUBDEV_ST(s, i, dev, DEV_UNDEFINED)
+
+/* dev: (struct rte_eth_dev *) fail-safe device */
+#define PREFERRED_SUBDEV(dev) \
+       (&PRIV(dev)->subs[0])
+
+/* dev: (struct rte_eth_dev *) fail-safe device */
+#define TX_SUBDEV(dev)                                                   \
+       (PRIV(dev)->subs_tx >= PRIV(dev)->subs_tail                ? NULL \
+        : (PRIV(dev)->subs[PRIV(dev)->subs_tx].state < DEV_PROBED ? NULL \
+        : &PRIV(dev)->subs[PRIV(dev)->subs_tx]))
+
+/**
+ * s:   (struct sub_device *)
+ * ops: (struct eth_dev_ops) member
+ */
+#define SUBOPS(s, ops) \
+       (ETH(s)->dev_ops->ops)
+
+#ifndef NDEBUG
+#include <stdio.h>
+#define DEBUG__(m, ...)                                                \
+       (fprintf(stderr, "%s:%d: %s(): " m "%c",                \
+                __FILE__, __LINE__, __func__, __VA_ARGS__),    \
+        (void)0)
+#define DEBUG_(...)                            \
+       (errno = ((int []){                     \
+               *(volatile int *)&errno,        \
+               (DEBUG__(__VA_ARGS__), 0)       \
+       })[0])
+#define DEBUG(...) DEBUG_(__VA_ARGS__, '\n')
+#define INFO(...) DEBUG(__VA_ARGS__)
+#define WARN(...) DEBUG(__VA_ARGS__)
+#define ERROR(...) DEBUG(__VA_ARGS__)
+#else
+#define DEBUG(...) ((void)0)
+#define LOG__(level, m, ...) \
+       RTE_LOG(level, PMD, "net_failsafe: " m "%c", __VA_ARGS__)
+#define LOG_(level, ...) LOG__(level, __VA_ARGS__, '\n')
+#define INFO(...) LOG_(INFO, __VA_ARGS__)
+#define WARN(...) LOG_(WARNING, "WARNING: " __VA_ARGS__)
+#define ERROR(...) LOG_(ERR, "ERROR: " __VA_ARGS__)
+#endif
+
+/* inlined functions */
+
+static inline uint8_t
+fs_find_next(struct rte_eth_dev *dev, uint8_t sid,
+               enum dev_state min_state)
+{
+       while (sid < PRIV(dev)->subs_tail) {
+               if (PRIV(dev)->subs[sid].state >= min_state)
+                       break;
+               sid++;
+       }
+       if (sid >= PRIV(dev)->subs_tail)
+               return PRIV(dev)->subs_tail;
+       return sid;
+}
+
+#endif /* _RTE_ETH_FAILSAFE_PRIVATE_H_ */

drivers/net/failsafe/failsafe_rxtx.c new file +107 -0

new file mode 100644
index 0000000..a45b4e5
@@ -0,0 +1,107 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright 2017 6WIND S.A.
+ *   Copyright 2017 Mellanox.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of 6WIND S.A. nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <rte_mbuf.h>
+#include <rte_ethdev.h>
+
+#include "failsafe_private.h"
+
+/*
+ * TODO: write fast version,
+ * without additional checks, to be activated once
+ * everything has been verified to comply.
+ */
+uint16_t
+failsafe_rx_burst(void *queue,
+                 struct rte_mbuf **rx_pkts,
+                 uint16_t nb_pkts)
+{
+       struct fs_priv *priv;
+       struct sub_device *sdev;
+       struct rxq *rxq;
+       void *sub_rxq;
+       uint16_t nb_rx;
+       uint8_t nb_polled, nb_subs;
+       uint8_t i;
+
+       rxq = queue;
+       priv = rxq->priv;
+       nb_subs = priv->subs_tail - priv->subs_head;
+       nb_polled = 0;
+       for (i = rxq->last_polled; nb_polled < nb_subs; nb_polled++) {
+               i++;
+               if (i == priv->subs_tail)
+                       i = priv->subs_head;
+               sdev = &priv->subs[i];
+               if (unlikely(ETH(sdev) == NULL))
+                       continue;
+               if (unlikely(ETH(sdev)->rx_pkt_burst == NULL))
+                       continue;
+               if (unlikely(sdev->state != DEV_STARTED))
+                       continue;
+               sub_rxq = ETH(sdev)->data->rx_queues[rxq->qid];
+               nb_rx = ETH(sdev)->
+                       rx_pkt_burst(sub_rxq, rx_pkts, nb_pkts);
+               if (nb_rx) {
+                       rxq->last_polled = i;
+                       return nb_rx;
+               }
+       }
+       return 0;
+}
+
+/*
+ * TODO: write fast version,
+ * without additional checks, to be activated once
+ * everything has been verified to comply.
+ */
+uint16_t
+failsafe_tx_burst(void *queue,
+                 struct rte_mbuf **tx_pkts,
+                 uint16_t nb_pkts)
+{
+       struct sub_device *sdev;
+       struct txq *txq;
+       void *sub_txq;
+
+       txq = queue;
+       sdev = TX_SUBDEV(txq->priv->dev);
+       if (unlikely(sdev == NULL))
+               return 0;
+       if (unlikely(ETH(sdev) == NULL))
+               return 0;
+       if (unlikely(ETH(sdev)->tx_pkt_burst == NULL))
+               return 0;
+       sub_txq = ETH(sdev)->data->tx_queues[txq->qid];
+       return ETH(sdev)->tx_pkt_burst(sub_txq, tx_pkts, nb_pkts);
+}

drivers/net/failsafe/rte_pmd_failsafe_version.map new file +4 -0

new file mode 100644
index 0000000..b6d2840
@@ -0,0 +1,4 @@
+DPDK_17.08 {
+
+       local: *;
+};

mk/rte.app.mk +1 -1

index dbd3614..d7581b7 100644
@@ -120,6 +120,7 @@ _LDLIBS-$(CONFIG_RTE_LIBRTE_E1000_PMD)      += -lrte_pmd_e1000
 _LDLIBS-$(CONFIG_RTE_LIBRTE_ENA_PMD)        += -lrte_pmd_ena
 _LDLIBS-$(CONFIG_RTE_LIBRTE_ENIC_PMD)       += -lrte_pmd_enic
 _LDLIBS-$(CONFIG_RTE_LIBRTE_FM10K_PMD)      += -lrte_pmd_fm10k
+_LDLIBS-$(CONFIG_RTE_LIBRTE_PMD_FAILSAFE)   += -lrte_pmd_failsafe
 _LDLIBS-$(CONFIG_RTE_LIBRTE_I40E_PMD)       += -lrte_pmd_i40e
 _LDLIBS-$(CONFIG_RTE_LIBRTE_IXGBE_PMD)      += -lrte_pmd_ixgbe
 ifeq ($(CONFIG_RTE_LIBRTE_KNI),y)
--

2.1.4

Quoted 11 lines · Gaetan Rivet

drivers/net/failsafe/failsafe_args.c new file +11 -0

+static int
+fs_parse_device(struct sub_device *sdev, char *args)
+{
+       struct rte_devargs *d;
+       int ret;
+
+       d = &sdev->devargs;
+       DEBUG("%s", args);
+       ret = rte_eal_devargs_parse(args, d);
+       if (ret) {
+               DEBUG("devargs parsing failed with code %d", ret);

drivers/net/failsafe/failsafe_args.c line 93

Quoted 4 lines

drivers/net/failsafe/failsafe_args.c new file +4 -0

+               return ret;
+       }
+       sdev->bus = d->bus;
+       sdev->state = DEV_PARSED;

You seem to be mostly interested in the bus name for the device. Why don't you track this via your sub_device structure instead of using rte_devargs?

Quoted 3 lines

drivers/net/failsafe/failsafe.c new file +3 -0

+       return 0;
+}
+

Quoted 7 lines

drivers/net/failsafe/failsafe_eal.c new file +7 -0

+static struct rte_eth_dev *
+fs_find_ethdev(const struct rte_device *dev)
+{
+       struct rte_eth_dev *eth_dev;
+       uint8_t i;
+
+       RTE_ETH_FOREACH_DEV(i) {

drivers/net/failsafe/failsafe_eal.c line 50

Quoted 6 lines

drivers/net/failsafe/failsafe_eal.c new file +6 -0

+               eth_dev = &rte_eth_devices[i];
+               if (eth_dev->device == dev)
+                       return eth_dev;
+       }
+       return NULL;
+}

Why don't you use rte_eth_dev_allocated() here?

Quoted 11 lines

drivers/net/failsafe/failsafe_eal.c new file +11 -0

+
+static int
+fs_bus_init(struct rte_eth_dev *dev)
+{
+       struct sub_device *sdev;
+       struct rte_device *rdev;
+       struct rte_devargs *da;
+       uint8_t i;
+       int ret;
+
+       FOREACH_SUBDEV(sdev, i, dev) {

drivers/net/failsafe/failsafe_eal.c line 67

Quoted 6 lines

drivers/net/failsafe/failsafe_eal.c new file +6 -0

+               if (sdev->state != DEV_PARSED)
+                       continue;
+               da = &sdev->devargs;
+               rdev = rte_eal_hotplug_add(da->bus->name,
+                                          da->name,
+                                          da->args);

Why don't you track the bus name through your sub_device structure instead?

drivers/net/failsafe/failsafe_args.c new file +0 -0

+#include <errno.h>

11/07/2017 01:19, Gaetan Rivet:

Quoted 2 lines

drivers/net/failsafe/failsafe_args.c new file +0 -0

[...]

Quoted 1 line

+#include <errno.h>

Do you really need errno in this file?

Quoted 7 lines

11/07/2017 01:19, Gaetan Rivet:

Quoted 2 lines

drivers/net/failsafe/failsafe_args.c new file +0 -0

[...]

Quoted 1 line

+#include <errno.h>

Do you really need errno in this file?

It's used for error handling on strtoull and popen. Why, is it a problem to include errno.h?

11/07/2017 23:48, Gaëtan Rivet:

Quoted 7 lines · Thomas Monjalon

Quoted 6 lines

11/07/2017 01:19, Gaetan Rivet:

Quoted 2 lines

drivers/net/failsafe/failsafe_args.c new file +0 -0

[...]

Quoted 1 line

+#include <errno.h>

Quoted 4 lines

Quoted 1 line

Do you really need errno in this file?

It's used for error handling on strtoull and popen. Why, is it a problem to include errno.h?

I don't see errno used in this patch.
I think it is used in next patches.