Release Notes

Package Version: 2.0

April 03, 2015

Contents

• 1. Description of Release
  • 1.1. Using DPDK Upgrade Patches
  • 1.2. Documentation Roadmap
• 2. New Features
• 3. Supported Features
• 4. Supported Operating Systems
• 5. Updating Applications from Previous Versions
  • 5.1. DPDK 1.7 to DPDK 1.8
  • 5.2. Intel® DPDK 1.6 to DPDK 1.7
  • 5.3. Intel® DPDK 1.5 to Intel® DPDK 1.6
  • 5.4. Intel® DPDK 1.4 to Intel® DPDK 1.5
  • 5.5. Intel® DPDK 1.3 to Intel® DPDK 1.4.x
  • 5.6. Intel® DPDK 1.2 to Intel® DPDK 1.3
  • 5.7. Intel® DPDK 1.1 to Intel® DPDK 1.2
• 6. Known Issues and Limitations
  • 6.1. Unit Test for Link Bonding may fail at test_tlb_tx_burst()
  • 6.2. Pause Frame Forwarding does not work properly on igb
  • 6.3. In packets provided by the PMD, some flags are missing
  • 6.4. The rte_malloc library is not fully implemented
  • 6.5. HPET reading is slow
  • 6.6. HPET timers do not work on the Osage customer reference platform
  • 6.7. Not all variants of supported NIC types have been used in testing
  • 6.8. Multi-process sample app requires exact memory mapping
  • 6.9. Packets are not sent by the 1 GbE/10 GbE SR-IOV driver when the source MAC address is not the MAC address assigned to the VF NIC
  • 6.10. SR-IOV drivers do not fully implement the rte_ethdev API
  • 6.11. PMD does not work with –no-huge EAL command line parameter
  • 6.12. Some hardware off-load functions are not supported by the VF Driver
  • 6.13. Kernel crash on IGB port unbinding
  • 6.14. Twinpond and Ironpond NICs do not report link status correctly
  • 6.15. Discrepancies between statistics reported by different NICs
  • 6.16. Error reported opening files on DPDK initialization
  • 6.17. Intel® QuickAssist Technology sample application does not work on a 32-bit OS on Shumway
  • 6.18. IEEE1588 support possibly not working with an Intel® Ethernet Controller I210 NIC
  • 6.19. Differences in how different Intel NICs handle maximum packet length for jumbo frame
  • 6.20. Binding PCI devices to igb_uio fails on Linux* kernel 3.9 when more than one device is used
  • 6.21. GCC might generate Intel® AVX instructions forprocessors without Intel® AVX support
  • 6.22. Ethertype filter could receive other packets (non-assigned) in Niantic
  • 6.23. Cannot set link speed on Intel® 40G ethernet controller
  • 6.24. Stopping the port does not down the link on Intel® 40G ethernet controller
  • 6.25. Devices bound to igb_uio with VT-d enabled do not work on Linux* kernel 3.15-3.17
• 7. Resolved Issues
  • 7.1. Running TestPMD with SRIOV in Domain U may cause it to hang when XENVIRT switch is on
  • 7.2. Vhost-xen cannot detect Domain U application exit on Xen version 4.0.1
  • 7.3. Virtio incorrect header length used if MSI-X is disabled by kernel driver
  • 7.4. Unstable system performance across application executions with 2MB pages
  • 7.5. Link status change not working with MSI interrupts
  • 7.6. KNI does not provide Ethtool support for all NICs supported by the Poll-Mode Drivers
  • 7.7. Linux IPv4 forwarding is not stable with vhost-switch on high packet rate
  • 7.8. PCAP library overwrites mbuf data before data is used
  • 7.9. MP Client Example app - flushing part of TX is not working for some ports if set specific port mask with skipped ports
  • 7.10. Packet truncation with Intel® I350 Gigabit Ethernet Controller
  • 7.11. Device initialization failure with Intel® Ethernet Server Adapter X520-T2
  • 7.12. DPDK kernel module is incompatible with Linux kernel version 3.3
  • 7.13. Initialization failure with Intel® Ethernet Controller X540-T2
  • 7.14. rte_eth_dev_stop() function does not bring down the link for 1 GB NIC ports
  • 7.15. It is not possible to adjust the duplex setting for 1GB NIC ports
  • 7.16. Calling rte_eth_dev_stop() on a port does not free all the mbufs in use by that port
  • 7.17. PMD does not always create rings that are properly aligned in memory
  • 7.18. Checksum offload might not work correctly when mixing VLAN-tagged and ordinary packets
  • 7.19. Port not found issue with Intel® 82580 Gigabit Ethernet Controller
  • 7.20. Packet mbufs may be leaked from mempool if rte_eth_dev_start() function fails
  • 7.21. Promiscuous mode for 82580 NICs can only be enabled after a call to rte_eth_dev_start for a port
  • 7.22. Incorrect CPU socket information reported in /proc/cpuinfo can prevent the DPDK from running
  • 7.23. L3FWD sample application may fail to transmit packets under extreme conditions
  • 7.24. L3FWD-VF might lose CRC bytes
  • 7.25. 32-bit DPDK sample applications fails when using more than one 1 GB hugepage
  • 7.26. l2fwd fails to launch if the NIC is the Intel® 82571EB Gigabit Ethernet Controller
  • 7.27. 32-bit DPDK applications may fail to initialize on 64-bit OS
  • 7.28. Lpm issue when using prefixes > 24
  • 7.29. IXGBE PMD hangs on port shutdown when not all packets have been sent
  • 7.30. Config file change can cause build to fail
  • 7.31. rte_cmdline library should not be used in production code due to limited testing
  • 7.32. Some *_INITIALIZER macros are not compatible with C++
  • 7.33. No traffic through bridge when using exception_path sample application
  • 7.34. Segmentation Fault in testpmd after config fails
  • 7.35. Linux kernel pci_cfg_access_lock() API can be prone to deadlock
  • 7.36. When running multi-process applications, “rte_malloc” functions cannot be used in secondary processes
  • 7.37. Configuring maximum packet length for IGB with VLAN enabled may not take intoaccount the length of VLAN tag
  • 7.38. Intel® I210 Ethernet controller always strips CRC of incoming packets
  • 7.39. EAL can silently reserve less memory than requested
  • 7.40. SSH connectivity with the board may be lost when starting a DPDK application
  • 7.41. Remote network connections lost when running autotests or sample applications
  • 7.42. KNI may not work properly in a multi-process environment
  • 7.43. Hash library cannot be used in multi-process applications with multiple binaries
  • 7.44. Unused hugepage files are not cleared after initialization
  • 7.45. Packet reception issues when virtualization is enabled
  • 7.46. Double VLAN does not work on Intel® 40GbE ethernet contoller
• 8. ABI policy
  • 8.1. Examples of Deprecation Notices
  • 8.2. Deprecation Notices
• 9. Frequently Asked Questions (FAQ)
  • 9.1. When running the test application, I get “EAL: map_all_hugepages(): open failed: Permission denied Cannot init memory”?
  • 9.2. If I want to change the number of TLB Hugepages allocated, how do I remove the original pages allocated?
  • 9.3. If I execute “l2fwd -c f -m 64 –n 3 – -p 3”, I get the following output, indicating that there are no socket 0 hugepages to allocate the mbuf and ring structures to?
  • 9.4. I am running a 32-bit DPDK application on a NUMA system, and sometimes the application initializes fine but cannot allocate memory. Why is that happening?
  • 9.5. On application startup, there is a lot of EAL information printed. Is there any way to reduce this?
  • 9.6. How can I tune my network application to achieve lower latency?
  • 9.7. Without NUMA enabled, my network throughput is low, why?
  • 9.8. I am getting errors about not being able to open files. Why?
  • 9.9. Does my kernel require patching to run the DPDK?
  • 9.10. VF driver for IXGBE devices cannot be initialized.
  • 9.11. Is it safe to add an entry to the hash table while running?
  • 9.12. What is the purpose of setting iommu=pt?
  • 9.13. When trying to send packets from an application to itself, meaning smac==dmac, using Intel(R) 82599 VF packets are lost.
  • 9.14. Can I split packet RX to use DPDK and have an application’s higher order functions continue using Linux* pthread?
  • 9.15. Is it possible to exchange data between DPDK processes and regular userspace processes via some shared memory or IPC mechanism?
  • 9.16. Can the multiple queues in Intel(R) I350 be used with DPDK?
  • 9.17. How can hugepage-backed memory be shared among multiple processes?