Also, make sure to start the actual text at the margin.
=========================================================
+* **Reorganized the mbuf structure.**
+
+ * Align fields to facilitate the writing of ``data_off``, ``refcnt``, and
+ ``nb_segs`` in one operation.
+ * Use 2 bytes for port and number of segments.
+ * Move the sequence number in the second cache line.
+ * Add a timestamp field.
+ * Set default value for ``refcnt``, ``next`` and ``nb_segs`` at mbuf free.
+
+* **Added mbuf raw free API**
+
+ Moved ``rte_mbuf_raw_free()`` and ``rte_pktmbuf_prefree_seg()`` functions to
+ the public API.
+
* **Added free Tx mbuf on demand API.**
Added a new function ``rte_eth_tx_done_cleanup()`` which allows an application
to request the driver to release mbufs from their Tx ring that are no longer
in use, independent of whether or not the ``tx_rs_thresh`` has been crossed.
+* **Added EAL dynamic log framework.**
+
+ Added new APIs to dynamically register named log types, and control
+ the level of each type independently.
+
+* **Added descriptor status ethdev API.**
+
+ Added a new API to get the status of a descriptor.
+
+ For Rx, it is almost similar to the ``rx_descriptor_done`` API, except
+ it differentiates descriptors which are hold by the driver and not
+ returned to the hardware. For Tx, it is a new API.
+
* **Increased number of next hops for LPM IPv6 to 2^21.**
The next_hop field is extended from 8 bits to 21 bits for IPv6.
+* **Added VFIO hotplug support.**
+
+ How hotplug supported with UIO and VFIO drivers.
+
* **Added powerpc support in pci probing for vfio-pci devices.**
sPAPR IOMMU based pci probing enabled for vfio-pci devices.
+* **Kept consistent PMD batching behaviour.**
+
+ Removed the limit of fm10k/i40e/ixgbe TX burst size and vhost RX/TX burst size
+ in order to support the same policy of "make an best effort to RX/TX pkts"
+ for PMDs.
+
+* **Updated the ixgbe base driver.**
+
+ Updated the ixgbe base driver, including the following changes:
+
+ * Add link block check for KR.
+ * Complete HW initialization even if SFP is not present.
+ * Add VF xcast promiscuous mode.
+
+* **Added powerpc support for i40e and its vector PMD .**
+
+ i40e PMD and its vector PMD enabled by default in powerpc.
+
+* **Added VF max bandwidth setting on i40e.**
+
+ i40e HW supports to set the max bandwidth for a VF. Enable this capability.
+
+* **Added VF TC min bandwidth setting on i40e.**
+
+ i40e HW supports to set the allocated bandwidth for a TC on a VF. Enable this
+ capability.
+
+* **Added VF TC max bandwidth setting on i40e.**
+
+ i40e HW supports to set the max bandwidth for a TC on a VF. Enable this
+ capability.
+
+* **Added TC strict priority mode setting on i40e.**
+
+ There're 2 TX scheduling modes supported for TCs by i40e HW, round ribon mode
+ and strict priority mode. By default it's round robin mode. Enable the
+ capability to change the TX scheduling mode for a TC. It's a global setting
+ on a physical port.
+
+* **Added i40e dynamic device personalization support.**
+
+ * Added dynamic device personalization processing to i40e FW.
+
+* **Added TSO support for tunneled and non-tunneled packets on mlx5 driver.**
+
+ Added support for Hardware TSO for tunneled and non-tunneled packets.
+ Tunneling protocols supported are GRE and VXLAN.
+
+* **Added support for Rx interrupts on mlx5 driver.**
+
+ Rx queues can be armed with an interrupt which will trigger on the
+ next packet arrival.
+
+* **Updated the sfc_efx driver.**
+
+ * Generic flow API support for Ethernet, VLAN, IPv4, IPv6, UDP and TCP
+ pattern items with QUEUE action for ingress traffic.
+
+ * Support virtual functions (VFs)
+
+* **Added LiquidIO network PMD.**
+
+ Added poll mode driver support for Cavium LiquidIO II server adapter VFs.
+
+* **Added support for the Wind River Systems AVP PMD.**
+
+ Added a new networking driver for the AVP device type. Theses devices are
+ specific to the Wind River Systems virtualization platforms.
+
+* **Added vmxnet3 version 3 support.**
+
+ Added support for vmxnet3 version 3 which includes several
+ performance enhancements viz. configurable TX data ring, Receive
+ Data Ring, ability to register memory regions.
+
+* **Updated the tap driver.**
+
+ * Support MTU modification.
+ * Support packet type for Rx.
+ * Support segmented packets on Rx and Tx.
+ * Speed up Rx on tap when no packets are available.
+ * Support capturing traffic from another netdevice.
+ * Dynamically change link status when the underlying interface state changes.
+ * Generic flow API support for Ethernet, VLAN, IPv4, IPv6, UDP and TCP pattern
+ items with DROP, QUEUE and PASSTHRU actions for ingress traffic.
+
+* **Added MTU feature support to Virtio and Vhost.**
+
+ Implemented new Virtio MTU feature into Vhost and Virtio:
+
+ * Add ``rte_vhost_mtu_get()`` API to Vhost library.
+ * Enable Vhost PMD's MTU get feature.
+ * Get max MTU value from host in Virtio PMD
+
+* **Added interrupt mode support for virtio-user.**
+
+ Implemented Rxq interrupt mode and LSC support for virtio-user as a virtual
+ device. Supported cases:
+
+ * Rxq interrupt for virtio-user + vhost-user as the backend.
+ * Rxq interrupt for virtio-user + vhost-kernel as the backend.
+ * LSC interrupt for virtio-user + vhost-user as the backend.
+
+* **Added event driven programming model library (rte_eventdev).**
+
+ This API introduces event driven programming model.
+
+ In a polling model, lcores poll ethdev ports and associated
+ rx queues directly to look for packet. In an event driven model,
+ by contrast, lcores call the scheduler that selects packets for
+ them based on programmer-specified criteria. Eventdev library
+ added support for event driven programming model, which offer
+ applications automatic multicore scaling, dynamic load balancing,
+ pipelining, packet ingress order maintenance and
+ synchronization services to simplify application packet processing.
+
+ By introducing event driven programming model, DPDK can support
+ both polling and event driven programming models for packet processing,
+ and applications are free to choose whatever model
+ (or combination of the two) that best suits their needs.
+
+* **Added Software Eventdev PMD.**
+
+ Added support for the software eventdev PMD. The software eventdev is a
+ software based scheduler device that implements the eventdev API. This
+ PMD allows an application to configure a pipeline using the eventdev
+ library, and run the scheduling workload on a CPU core.
+
+* **Added Cavium OCTEONTX Eventdev PMD.**
+
+ Added the new octeontx ssovf eventdev driver for OCTEONTX devices. See the
+ "Event Device Drivers" document for more details on this new driver.
+
Resolved Issues
---------------
Also, make sure to start the actual text at the margin.
=========================================================
+* **LSC interrupt cannot work for virtio-user + vhost-kernel.**
+
+ LSC interrupt cannot be detected when setting the backend, tap device,
+ up/down as we fail to find a way to monitor such event.
+
API Changes
-----------
* removed the build-time setting ``CONFIG_RTE_RING_PAUSE_REP_COUNT``
* removed the function ``rte_ring_set_water_mark`` as part of a general
removal of watermarks support in the library.
+ * added an extra parameter to the burst/bulk enqueue functions to
+ return the number of free spaces in the ring after enqueue. This can
+ be used by an application to implement its own watermark functionality.
+ * added an extra parameter to the burst/bulk dequeue functions to return
+ the number elements remaining in the ring after dequeue.
* changed the return value of the enqueue and dequeue bulk functions to
match that of the burst equivalents. In all cases, ring functions which
operate on multiple packets now return the number of elements enqueued
- ``rte_ring_sc_dequeue_bulk``
- ``rte_ring_dequeue_bulk``
+ NOTE: the above functions all have different parameters as well as
+ different return values, due to the other listed changes above. This
+ means that all instances of the functions in existing code will be
+ flagged by the compiler. The return value usage should be checked
+ while fixing the compiler error due to the extra parameter.
+
+* **Reworked rte_vhost library**
+
+ The rte_vhost library has been reworked to make it generic enough so that
+ user could build other vhost-user drivers on top of it. To achieve that,
+ following changes have been made:
+
+ * The following vhost-pmd APIs are removed:
+
+ * ``rte_eth_vhost_feature_disable``
+ * ``rte_eth_vhost_feature_enable``
+ * ``rte_eth_vhost_feature_get``
+
+ * The vhost API ``rte_vhost_driver_callback_register(ops)`` is reworked to
+ be per vhost-user socket file. Thus, it takes one more argument:
+ ``rte_vhost_driver_callback_register(path, ops)``.
+
+ * The vhost API ``rte_vhost_get_queue_num`` is deprecated, instead,
+ ``rte_vhost_get_vring_num`` should be used.
+
+ * Following macros are removed in ``rte_virtio_net.h``
+
+ * ``VIRTIO_RXQ``
+ * ``VIRTIO_TXQ``
+ * ``VIRTIO_QNUM``
+
+ * Following net specific header files are removed in ``rte_virtio_net.h``
+
+ * ``linux/virtio_net.h``
+ * ``sys/socket.h``
+ * ``linux/if.h``
+ * ``rte_ether.h``
+
+ * The vhost struct ``virtio_net_device_ops`` is renamed to
+ ``vhost_device_ops``
+
+ * The vhost API ``rte_vhost_driver_session_start`` is removed. Instead,
+ ``rte_vhost_driver_start`` should be used, and no need to create a
+ thread to call it.
+
+ * The vhost public header file ``rte_virtio_net.h`` is renamed to
+ ``rte_vhost.h``
+
+
ABI Changes
-----------
Also, make sure to start the actual text at the margin.
=========================================================
+* **Reorganized the mbuf structure.**
+
+ The order and size of the fields in the ``mbuf`` structure changed,
+ as described in the `New Features`_ section.
+
Removed Items
-------------
librte_kni.so.2
librte_kvargs.so.1
librte_lpm.so.2
- librte_mbuf.so.2
+ + librte_mbuf.so.3
librte_mempool.so.2
librte_meter.so.1
librte_net.so.1
git.droids-corp.org / dpdk.git / blobdiff