summaryrefslogtreecommitdiffstats
path: root/src/vnet
AgeCommit message (Collapse)AuthorFilesLines
2021-10-20ip6: fix IPv6 address calculation error using "ip route add" CLIJieqiang Wang3-20/+61
Using VPP CLI "ip route add" to add static IPv6 entries outputs wrong results. Fix this error by correctly calculating IPv6 addresses with different increased ranges and grouping ip4/ip6 prefix calculation functionality into two functions. Type: fix Signed-off-by: Jieqiang Wang <jieqiang.wang@arm.com> Reviewed-by: Lijian Zhang <lijian.zhang@arm.com> Reviewed-by: Tianyu Li <tianyu.li@arm.com> Change-Id: If954876301ca2095f9331799a086f75db936f246
2021-10-20gso: add support for gso perf unittestMohsin Kazmi2-0/+13
Type: test Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com> Change-Id: I5d44a6ea24e4aa0842024a0961f1fb22c6e6419a
2021-10-19bfd: fix bfd_key_id updateAlexander Skorichenko1-4/+8
Type: fix Currently, auth activation CLI command "bfd upd session auth activate ... conf-key-id <cki> bfd-key-id <bki>" allows to change both key-ids to new values at once. But if only bfd-key-id should be corrected, e.g. as a result of mistyping, we can't do that in a single operation, and have to deactivate auth first and then reactivate it with a correctly entered pair of ids. Currently, backend's bfd_auth_activate() function returns immediately, with no action, if it finds that submitted conf-key-id matches the current record. No check on bfd-key-id value is made. With this fix, bfd_auth_activate() checks if session's bfd-key-id has to be changed to a new value, and if so, it updates and logs appropriately. Change-Id: I3b915a936cb1721707860bb503f70e7dd29e0ddd Signed-off-by: Alexander Skorichenko <askorichenko@netgate.com>
2021-10-19bfd: fix desired_min_tx_nsec updateAlexander Skorichenko1-0/+2
Type: fix Currently, the change in desired_min_tx value initiated locally doesn't result in actual update of the transmission interval. bfd_session_t structure has two fields for each of protocol's local time parameters desired_min_tx, required_min_rx In a case of a parameter update, "config_" prefixed fields store new values sent to remote bs.config_desired_min_tx_nsec, bs.config_required_min_rx_nsec Those prefixed "effective_", keep old values, still in charge, until new ones are being negotiated between peers: bs.effective_desired_min_tx_nsec, bs.effective_required_min_rx_nsec Currently, upon termination of the Poll Sequence (negotiation) only bfd_set_effective_required_min_rx() is called to update effective RX value. TX value remains unchanged. With this fix, add a call to bfd_set_effective_desired_min_tx() to set new, acknowledged TX value in effect. Signed-off-by: Alexander Skorichenko <askorichenko@netgate.com> Change-Id: I80b6746533839c9572598f1ad9dabb33e621a525
2021-10-18api: remove legacy pd msg handlersKlement Sekera2-2/+0
While looking at a coverity warning it was discovered that it's caused by return value modified by pd msg handlers. As these are legacy, it's a good time to remove them and also fix the warning this way. Type: fix Signed-off-by: Klement Sekera <ksekera@cisco.com> Change-Id: Ic72ab8b2b7a2e55188d1c31cfd18a74b7cf82c43
2021-10-18interface: add api test fileFilip Tehlar3-15/+1179
Type: improvement Signed-off-by: Filip Tehlar <ftehlar@cisco.com> Change-Id: Ib07029204ecf12bf2adb5a39afa54bc98fb81f34
2021-10-16interface: fix offload flag checkNathan Skrzypczak1-1/+4
We should compute offloads in interface_ouput.c in the case when any of the UDP, TCP or IP offload flags is not present in the interface capabilities. e.g. if the interface supports IP cksum offload but not tcp/udp we should still compute cksums here. Type: fix Change-Id: Ibaa3a56dbc92938dca45311c38f079d040052194 Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-10-14ip6-nd: fix coverity warningKlement Sekera1-0/+4
Add a missing return statement in case there are no parameters supplied to 'set ip6 nd proxy' CLI to avoid calling code with uninitialised parameters. Type: fix Signed-off-by: Klement Sekera <ksekera@cisco.com> Change-Id: Ie4fdb4df5d1af49471c421e5e7a6c2f885d2e8d2
2021-10-13docs: convert vpp doc md->rstNathan Skrzypczak20-1268/+1708
Type: improvement Change-Id: If453321785b04f9c16e8cea36fb1910efaeb2c59 Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-10-13session: app name should format with %vXiaoming Jiang1-1/+1
Type: fix Signed-off-by: Xiaoming Jiang <jiangxiaoming@outlook.com> Change-Id: I2c77066cc9f1d3063373cc9559cc5b369906cc24
2021-10-13interface: handle error during admin-up correctlyMatthew Smith1-3/+0
Type: fix In vnet_sw_interface_set_flags_helper(), the variable old_flags is set to the original value of vnet_sw_interface_t.flags for an interface. If an error occurs during the process of bringing an interface up, old_flags is used to restore the original value. Before the dev class or hw class admin_up_down_function can be called, but after modifying vnet_sw_interface_t.flags to it's new value, old_flags is set to the value of vnet_sw_interface_t.flags a second time. This discards the original flags that were being preserved. As a result, if an interface is being brought up and the dev class or hw class function fails, at the end VPP believes that interface is up. This can cause a crash if packets are routed through the interface and some RX/TX initialization was not completed because of the error while bringing the interface up. Change-Id: Ica6b6bac13c24e88c4136bf084cd392e6217e7d9 Signed-off-by: Matthew Smith <mgsmith@netgate.com>
2021-10-12fib: fix the drop counter for ipv6 RPF failuresNeale Ranns2-24/+30
Type: fix the only change to the mfib forwarding node is to set the error code, the rest is checkstyle formatting. The traces previously showed some bogus reason: 00:04:27:325550: ip6-mfib-forward-rpf entry 10 itf -1 flags 00:04:27:325551: ip6-drop fib:0 adj:10 flow:0 UDP: fe80::b203:eaff:fe02:604 -> ff02::1:2 tos 0x00, flow label 0x651ed, hop limit 1, payload length 64 UDP: 546 -> 547 length 64, checksum 0xec9a 00:04:27:325551: error-drop rx:GigabitEthernet6/0/0 00:04:27:325553: drop ip6-input: drops due to concurrent reassemblies limit Signed-off-by: Neale Ranns <neale@graphiant.com> Change-Id: I294684c36edc346b4ebdd83ba66888b3b2197704
2021-10-12ipsec: Performance improvement of ipsec4_output_node using flow cacheGovindarajan Mohandoss7-17/+264
Adding flow cache support to improve outbound IPv4/IPSec SPD lookup performance. Details about flow cache: Mechanism: 1. First packet of a flow will undergo linear search in SPD table. Once a policy match is found, a new entry will be added into the flow cache. From 2nd packet onwards, the policy lookup will happen in flow cache. 2. The flow cache is implemented using bihash without collision handling. This will avoid the logic to age out or recycle the old flows in flow cache. Whenever a collision occurs, old entry will be overwritten by the new entry. Worst case is when all the 256 packets in a batch result in collision and fall back to linear search. Average and best case will be O(1). 3. The size of flow cache is fixed and decided based on the number of flows to be supported. The default is set to 1 million flows. This can be made as a configurable option as a next step. 4. Whenever a SPD rule is added/deleted by the control plane, the flow cache entries will be completely deleted (reset) in the control plane. The assumption here is that SPD rule add/del is not a frequent operation from control plane. Flow cache reset is done, by putting the data plane in fall back mode, to bypass flow cache and do linear search till the SPD rule add/delete operation is complete. Once the rule is successfully added/deleted, the data plane will be allowed to make use of the flow cache. The flow cache will be reset only after flushing out the inflight packets from all the worker cores using vlib_worker_wait_one_loop(). Details about bihash usage: 1. A new bihash template (16_8) is added to support IPv4 5 tuple. BIHASH_KVP_PER_PAGE and BIHASH_KVP_AT_BUCKET_LEVEL are set to 1 in the new template. It means only one KVP is supported per bucket. 2. Collision handling is avoided by calling BV (clib_bihash_add_or_overwrite_stale) function. Through the stale callback function pointer, the KVP entry will be overwritten during collision. 3. Flow cache reset is done using BV (clib_bihash_foreach_key_value_pair) function. Through the callback function pointer, the KVP value is reset to ~0ULL. MRR performance numbers with 1 core, 1 ESP Tunnel, null-encrypt, 64B for different SPD policy matching indices: SPD Policy index : 1 10 100 1000 Throughput : MPPS/MPPS MPPS/MPPS MPPS/MPPS KPPS/MPPS (Baseline/Optimized) ARM Neoverse N1 : 5.2/4.84 4.55/4.84 2.11/4.84 329.5/4.84 ARM TX2 : 2.81/2.6 2.51/2.6 1.27/2.6 176.62/2.6 INTEL SKX : 4.93/4.48 4.29/4.46 2.05/4.48 336.79/4.47 Next Steps: Following can be made as a configurable option through startup conf at IPSec level: 1. Enable/Disable Flow cache. 2. Bihash configuration like number of buckets and memory size. 3. Dual/Quad loop unroll can be applied around bihash to further improve the performance. 4. The same flow cache logic can be applied for IPv6 as well as in IPSec inbound direction. A deeper and wider flow cache using bihash_40_8 can replace existing bihash_16_8, to make it common for both IPv4 and IPv6 in both outbound and inbound directions. Following changes are made based on the review comments: 1. ON/OFF flow cache through startup conf. Default: OFF 2. Flow cache stale entry detection using epoch counter. 3. Avoid host order endianness conversion during flow cache lookup. 4. Move IPSec startup conf to a common file. 5. Added SPD flow cache unit test case 6. Replaced bihash with vectors to implement flow cache. 7. ipsec_add_del_policy API is not mpsafe. Cleaned up inflight packets check in control plane. Type: improvement Signed-off-by: mgovind <govindarajan.Mohandoss@arm.com> Signed-off-by: Zachary Leaf <zachary.leaf@arm.com> Tested-by: Jieqiang Wang <jieqiang.wang@arm.com> Change-Id: I62b4d6625fbc6caf292427a5d2046aa5672b2006
2021-10-12session: set actual lcl ip on accepted ctFlorin Coras1-1/+1
Type: improvement Signed-off-by: Florin Coras <fcoras@cisco.com> Change-Id: Ic2ba5fa234a394acb524b61573fc49f2d58c2dea
2021-10-11ip: fix fib and mfib locksNathan Skrzypczak9-84/+92
This patches fixes an issue that could cause fib locks to underflow: if an API user deletes a fib and quickly recreates it, the fib may not have been actually deleted. As a result, the lock would not be incremented on the create call leading to the fib potentially disappearing afterwards - or to the lock to underflow when the fib is deleted again. In order to keep the existing API semantics, we use the locks with API and CLI source as flags. This means we need to use a different counter for the interface-related locks. This also prevents an issue where an interface being bound to a vrf via API and released via CLI could mess up the lock counter. Finally, this will help with cleaning up the interface-related locks on interface deletion in a later patch. Type: fix Change-Id: I93030a7660646d6dd179ddf27fe4e708aa11b90e Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com> Signed-off-by: Aloys Augustin <aloaugus@cisco.com>
2021-10-11docs: nitfixes in FEATURE.yamlNathan Skrzypczak2-2/+2
Type: improvement Change-Id: Iec585880085b12b08594a0640822cd831455d594 Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-10-08vxlan-gpe: add udp-port configuration supportArtem Glazychev5-280/+446
similar behavior as here: 839dcc0fb7313638d9b8f52a9db81350dddfe461 Type: improvement Signed-off-by: Artem Glazychev <artem.glazychev@xored.com> Change-Id: I1b0a8f8f3dab48839e27df7065cf5f786cf0b5e9
2021-10-07ipsec: fix protect update log if nexthop is NULLBenoît Ganne1-2/+3
If logging is on, it will try to print the address nh. Make sure it is not NULL. Type: fix Change-Id: I81c0295865901406d86e0d822a103b4d5adffe47 Signed-off-by: Benoît Ganne <bganne@cisco.com>
2021-10-07vppinfra: asan: improve overflow semanticBenoît Ganne1-2/+2
Type: improvement Change-Id: Ia63899b82e34f179f9efa921e4630b598f2a86cb Signed-off-by: Benoît Ganne <bganne@cisco.com>
2021-10-07udp: fix severity error infoFilip Tehlar4-25/+25
Type: fix Signed-off-by: Filip Tehlar <ftehlar@cisco.com> Change-Id: I415d68b39ecac546b531f6eb98bca51e7eb6f7f7
2021-10-07ip: fix path MTU node errors definitionBenoît Ganne3-3/+7
The path mtu node uses errors defined by ip fragmentation. Type: fix Change-Id: I1f173955919a4f555ab0309cd8201ec342a0ae92 Signed-off-by: Benoît Ganne <bganne@cisco.com>
2021-10-07arp: source address selectionEd Warnicke2-5/+13
https://gerrit.fd.io/r/c/vpp/+/30197 introduced SAS and inadvertently broke ping in a variety of situations: https://jira.fd.io/browse/VPP-1992 https://jira.fd.io/browse/VPP-1970 https://lists.fd.io/g/vpp-dev/topic/84038840 all of which seem to be rooted in situations where there's literally nothing smarter ping can do for source address selection than to pick the first IP on the interface. This can happen for: 1. P2P interfaces, see attempted fix: https://gerrit.fd.io/r/c/vpp/+/32801 2. Interfaces with /32 IP addresses intentionally assigned After some discussion, this problem was partially fixed in https://gerrit.fd.io/r/c/vpp/+/33449 Unforunately, while source selection was fixed in ping, it continued to be broken in arp/nd. This gerrit builds on https://gerrit.fd.io/r/c/vpp/+/33449 and fixes arp/nd. Type: fix Ticket: VPP-1970 Ticket: VPP-1992 Fixes: e2fe097424fb169dfe01421ff17b8ccd0c26b4a6 Change-Id: Ief60c321676a15f4f30bf4cd84d50b2f1efec432 Signed-off-by: Ed Warnicke <hagbard@gmail.com>
2021-10-06ip: check if interface has link-local address (addition)Artem Glazychev1-1/+6
previous - b31fbc47f5fcf8234c757558d7b0285348774086 Type: fix Signed-off-by: Artem Glazychev <artem.glazychev@xored.com> Change-Id: I7ea2d693d3ad5bf41ece066b3511fbfa156c1e4b
2021-10-06session: fix severity infoFilip Tehlar1-11/+10
Type: fix Signed-off-by: Filip Tehlar <ftehlar@cisco.com> Change-Id: I6548274f8c0ae2a183b1d221cb195de445c2819f
2021-10-06tcp: fix severity infoFilip Tehlar5-66/+66
Type: fix Signed-off-by: Filip Tehlar <ftehlar@cisco.com> Signed-off-by: Florin Coras <fcoras@cisco.com> Change-Id: Ibe39bc045c3b154209a83b59ef95a37c61b32c0c
2021-10-06docs: more nitfixesNathan Skrzypczak3-38/+42
Type: fix Change-Id: I41455e1cdc62e7c0baa148630b0701b042f3b156 Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-10-06docs: vnet comment nitfixesNathan Skrzypczak15-84/+108
Type: improvement Change-Id: Iac01d7830b53819ace8f199554be10ab89ecdb97 Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-10-06ip: add classifier-based ACLs support on ip puntBenoît Ganne5-291/+301
This feature allows one to add classifier-based ACLs on packets punted from the ip infra, eg. to only whitelist specific sender(s). Type: feature Change-Id: Idab37b188583efbca980038875fc3e540cb2e880 Signed-off-by: Benoît Ganne <bganne@cisco.com>
2021-10-05session: Add session_sapi_enable_disableNathan Skrzypczak5-5/+33
Type: feature This adds an API message to do the switch at runtime. Change-Id: Ice6b69c57f0bfbf5668182e25593362ff4133615 Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-10-05tap: free the tap_fds vec on interface deletionMohsin Kazmi1-0/+1
Type: fix Tap fds are stored in vector array but deleting tap was not freeing this vector. This patch fixes it. Change-Id: I5228e3b9f432c69cf2656b2ee7402360d775964b Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
2021-10-04interface: free the output_node_thread_runtimesMohsin Kazmi1-0/+1
Type: fix output_node_thread_runtimes was not freed when an interface is deleted. This patch fixes it. Change-Id: I763b0109be1904d43839528a346f3b9aa8927205 Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
2021-10-04interface: remove the redundant vec_free on rx_queue_indicesMohsin Kazmi1-1/+0
Type: fix vnet_delete_hw_interface() calls vec_free on rx_queue_indices. function vnet_hw_if_unregister_all_rx_queues() is used to free rx_queue_indices which is also called by vnet_delete_hw_interface(). So, second vec_free is redundant. Change-Id: Ibda4be38fd122d33532bb384c97b0b9e5f441134 Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
2021-10-04virtio: remove control queue support from virtio_show() for tap/tunMohsin Kazmi1-6/+0
Type: fix Tap/Tun interfaces do not have control queue. This patch removes the support of control queue from virtio_show() which is used by show tap/tun cli. Change-Id: Ib89144ad488ed548fb1ce50ee232a1b8659ccf29 Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
2021-10-04fib: fix unitialized padding in fib_api_next_hop_decodeBenoît Ganne1-4/+2
If the type is IPv4, makes sure the padding bytes are set to 0 as this is used by ip46_address_is_ip4() to detect the type. Type: fix Change-Id: I6a81fa05a6b227086853901bf3dcdc66e6d04d2c Signed-off-by: Benoît Ganne <bganne@cisco.com>
2021-10-04ip: fix punt for ipv6Benoît Ganne1-4/+9
Type: fix Change-Id: I583c30e9b63c0b0b6cd5fef0b2cb9ed7ec9856e2 Signed-off-by: Benoît Ganne <bganne@cisco.com>
2021-10-04interface: remove the input_node_thread_index_by_queueMohsin Kazmi2-4/+0
Type: fix input_node_thread_index_by_queue is not being used anymore. Change-Id: I0141fa0d024affb39771acf7516e064c5c8acfe9 Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
2021-10-03mpls: Save the L3 header offset in the meta-data before label impositionNeale Ranns1-1/+9
Type: improvement Subsequent features in the data-path can thus easily find the l3 header without parsing the label stack. Signed-off-by: Neale Ranns <neale@graphiant.com> Change-Id: I26f7d4bbe9186aeb8654706579c72424e8ecca2c
2021-10-01devices: add support for pseudo header checksumMohsin Kazmi5-15/+104
Type: improvement Linux uses pseudo header checksum when checksum of l4 is offloaded. This patch adds similar support in virtual interfaces. Change-Id: I6a94d1104e59356f95057e7c122e3be9cd8659a3 Signed-off-by: Aloys Augustin <aloaugus@cisco.com> Signed-off-by: Mohsin Kazmi <sykazmi@cisco.com>
2021-09-30fib: doc nitfixesNathan Skrzypczak10-19/+19
Type: improvement Change-Id: I29346c849a5e1ff3c2ea399671f9f50d075e9f18 Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-09-29classify: fix message IDs on API repliesMatthew Smith1-33/+42
Type: fix When the API cleanup of classify messages was done, the code was not updated to add the message enums to REPLY_MSG_ID_BASE. So the wrong message IDs are being sent back in replies to classify API requests. Add REPLY_MSG_ID_BASE when populated vl_msg_id on a reply. Change-Id: Ic7c828f14d42a346fc58fc9ff062b954f494cdbd Signed-off-by: Matthew Smith <mgsmith@netgate.com>
2021-09-29ipsec: Record the number of packets lost from an SANeale Ranns6-22/+88
Type: feature Gaps in the sequence numbers received on an SA indicate packets that were lost. Gaps are identified using the anti-replay window that records the sequences seen. Publish the number of lost packets in the stats segment at /net/ipsec/sa/lost Signed-off-by: Neale Ranns <neale@graphiant.com> Change-Id: I8af1c09b7b25a705e18bf82e1623b3ce19e5a74d
2021-09-29tap: Fix tap create with nsNathan Skrzypczak1-55/+24
This fixes the interface creation passing a netns. [0] made the renaming of the new tuntap interface before switching netns Thus, preventing creating an interface in another netns if one exists in VPP's netns with the same name. This also fixes restore netns on errors Type: fix [0] https://gerrit.fd.io/r/c/vpp/+/33696 Change-Id: I5c83bb37d664057bcf231cd0c636f0e51aa542ad Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-09-29fib: add barrier lock for fib_urpf_list_pool and fib_entry_poolStanislav Zaikin2-0/+22
Pools fib_urpf_list_pool and fib_entry_pool can grow while ip6_urpf_loose_check/fib_entry_get_flags_for_source are being executed. That may result as a crash in mt environment. Type: fix Change-Id: I44ca2cb70255e7aaf2e1f7a7d2eecd25cbdd0aaa Signed-off-by: Stanislav Zaikin <zstaseg@gmail.com>
2021-09-28api: API trace improvementsFilip Tehlar1-14/+17
Type: improvement * add support for JSON format in API trace * add ability to replay JSON API trace in both VPP and VAT2 * use CRC for backward compatibility check during JSON API replay * fix API trace CLI (and remove duplicits) * remove custom dump * remove vppapitrace.py * update docs accordingly Change-Id: I5294f68bebe6cbe738630f457f3a87720e06486b Signed-off-by: Filip Tehlar <ftehlar@cisco.com> Signed-off-by: Ole Troan <ot@cisco.com>
2021-09-28misc: vpe.api messages dynamically allocatedOle Troan1-0/+9
This is the last in the series of moving API messages from vpp/api/vpe.api to vlibmemory/memclnt.api. This patch makes the remaining vpe.api messages dynamic, to help VAT2 binary-api command. Moves the VAT test code to a separate file and removes the now unnused API meta files. Type: improvement Signed-off-by: Ole Troan <ot@cisco.com> Change-Id: I01dd78eaff1d3715dff17d2643bf0f7f0089935b Signed-off-by: Ole Troan <ot@cisco.com>
2021-09-28devices: Add queues params in create_ifNathan Skrzypczak7-96/+185
Type: feature Change-Id: I027ff2c5c905a7ccebd3705a58e35218a94f4880 Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-09-28sr: add API test filesFilip Tehlar3-0/+370
Type: improvement Signed-off-by: Filip Tehlar <ftehlar@cisco.com> Change-Id: Iefc88107ae96915570ae425a527c3969f7ce7b1d
2021-09-27fib: fix crash on exporter tracker removeVladislav Grishenko1-4/+3
Exported entries are tracked only when the prefix found in the export FIB is really attached, exporter tracker is not set if the export entry is not valid for export, ex. for special FIB entries - default route, zeronet, mcast and broadcast prefixes. When imported entries need to be purged, such unset exporter tracker is being removed by non-initialized index with absent delegate entries, causing corresponding assert and crash. Type: fix Signed-off-by: Vladislav Grishenko <themiron@yandex-team.ru> Change-Id: Ib24a2e7853a03a960577872480213e1e8097da5a
2021-09-27virtio: Still init unused txqNathan Skrzypczak1-17/+0
Type: fix For some virtio backends, packets don't seem to flow if all the queues that have been requested are not initialized Change-Id: I41765d668497ff954e6d69d36836590947fd93b1 Signed-off-by: Nathan Skrzypczak <nathan.skrzypczak@gmail.com>
2021-09-27gso: do not gro on small packetsAloys Augustin1-3/+41
This has two benefits: fix an issue where small packets can be padded to 64 bytes by the ethernet layer, and we included these bytes in the tcp stream; and also if we receive a small packet, it signals that this tcp stream is likely more latency-sensitive than throughput- sensitive, and skipping gro buffering should be beneficial. Type: fix Change-Id: I165b97673d8cdce95ebf0a66c362ae9f9e6f3f34 Signed-off-by: Aloys Augustin <aloaugus@cisco.com>