The previous fix was adding the sentinel before refilling rx,
which gave the NIC time to overwrite it with a new descriptor.
Ticket: VPP-2087
Type: fix
Fixes: 8b4d474abd62c623502ad9a4a279a9b4535ca0c1
Change-Id: I32bde4a763a62fb66c5c3871d9f10af6066e2d47
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
Make sure ct client segment handles do not collide if multi worker
application establishes cut-through sessions to only one server segment
manager.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I905379f9ed73c64d57a826a3e97d53dab3a87517
The previous change made CSIT virtio tests fail,
but those tests are not part of trending.
Ticket: VPP-2088
Type: fix
Fixes: a181eaa59bb2ff2784376918e95bbf92e5340db1
Change-Id: If0439a030c051894e07007da9cf0a2e4dc1434c3
Signed-off-by: Vratko Polak <vrpolak@cisco.com>
This is because mlx5_pci is also compatible with another
series of NICs such as ConnectX-5 and ConnectX-6.
Type: fix
Change-Id: I10f0468bbe36ab61c72fb3dc0aa898f8e2f9e88c
Signed-off-by: Nobuhiro MIKI <nmiki@yahoo-corp.jp>
Type: fix
In esp_encrypt_inline(), an index and pointer to the last processed SA
are stored. If the next packet uses the same SA, we defer on updating
counters until a different SA is encountered.
The pointer was being retrieved, then the SA was checked to see if the
packet should be dropped due to no crypto/integ algs, then the index was
updated. If the check failed, we would skip further processing and now
the pointer refers to a different SA than the index. When you have a
batch of packets that are encrypted using an SA followed by a packet
which is dropped for no algs and then more packets to be encrypted using
the original SA, the packets that arrive after the one that was dropped
end up being processed using a pointer that refers to the wrong SA data.
This can result in a segv.
Update the current_sa_index at the same time that the sa0 pointer is
updated.
Signed-off-by: Matthew Smith <mgsmith@netgate.com>
Change-Id: I65f1511a37475b4f737f5e1b51749c0a30e88806
DHCPv4 aka BOOTP is somewhat awkward. A DHCP client
on an interface must receive DHCP packets to
the broadcast address or to a unicast address.
Apparently before it's been assigned to itself.
Add this new API to allow external DHCP clients
enable the DHCP client detect feature per interface.
Type: improvement
Change-Id: If55aac03f25a045496be483940e4f5e7e18885b9
Signed-off-by: Ole Troan <otroan@employees.org>
The recent TX flows generation fix introduced "l3_hdr_offset" which
represents the offset of the IP header in the buffer's data. The problem
is that it is erroneously defined as a 16-bit unsigned integer. If the
calculated offset is negative, "l3_hdr_offset" will get a value close to
UINT16_MAX. And the code will search the IP header somewhere beyond the
buffer's data. For example, this will occur in the case when an ICMP
error is being sent in response to a received packet.
With this fix, make "l3_hdr_offset" a signed integer.
Type: fix
Change-Id: I6f1283c7ba02656d0f592519b5863e68348c5583
Signed-off-by: Alexander Chernavin <achernavin@netgate.com>
Currently, when IPFIX records generation is enabled for an interface in
the TX direction, some rewritten traffic is being sent from that
interface, and the Ethernet header's location has changed due to
rewriting, generated TX flows will contain fields with wrong and zero
values. For example, that can be observed when traffic is rewritten from
a subinterface to a hardware interface (i.e. when tags are removed). A
TX flow generated in this case will have wrong L2 fields because of an
incorrectly located Ethernet header. And zero L3/L4 fields because the
Ethernet type will match neither IP4 nor IP6.
The same code is executed to generate flows for both input and output
features. And the same mechanism is applied to identify the Ethernet
header in the buffer's data. However, such general code usually works
with the buffer's data conditionally based on the direction. For most
input features, the buffer's current_data will likely point to the IP
header. For most output features, the buffer's current_data will likely
point to the Ethernet header.
With this fix:
- Keep relying on ethernet_buffer_get_header() to locate the Ethernet
header for input features. And start using vlib_buffer_get_current()
to locate the Ethernet header for output features. The function will
account for the Ethernet header's position change in the buffer's
data if there is rewriting.
- After fixing Ethernet header determination in the buffer's data,
L3/L4 fields will contain non-zero but still incorrect data. That is
because IP header determination needs to be fixed too. It currently
relies on the fact that the Ethernet header is always located at the
beginning of the buffer's data and that l2_hdr_sz can be used as an
IP header offset. However, this may not be the case after rewriting.
So start calculating the actual offset of the IP header in the
buffer's data.
- Add a unit test to cover the case.
Type: fix
Change-Id: Icf3f9e6518912d06dff0d5aa48e103b3dc94edb7
Signed-off-by: Alexander Chernavin <achernavin@netgate.com>
Do not force cleanup of tcp half-open connection if tcp's cleanup
notification to tls is pending.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I7bccbe8429a4aab10df1c89b66138b967e04ac19
Updated the gtpu plugin code to support the PDU Session user plane protocol, required for 5G, as
specified in 3GPP TS 38.415 version 17.0.0. This enables some initial support of 5G gNodeB's with
the gtpu plugin.
New features:
- Basic support for the GTP-U Extension Header Flag.
Packets with one extension can now be decapsulated.
This enables basic support of the PDU Session user plane protocol (3GPP TS 38.415 version 17.0.0).
New tunnels can be created with a PDU enable flag and a 6-bit QoS Flow Identifier (QFI).
With this, encapsulated packets will have the PDU Session extension header, and the QFI set.
- Ability to forward GTP-U packets that are not handled by the plugin directly.
Only GTP-U packets with a message type of 255 (G-PDU) are handled directly.
However, 3GPP TS 29.281 defines several other message types like echo and error indication.
A new feature is added to optionally forward unknown or unsupported packets to a new IP address.
This works separately for unknown GTP-U message types, unknown TEIDs, and packets with an unknown
GTP-U header.
This allows both echo and error indications from a 5G gNodeB to be handled by a different system
outside VPP.
- Simple way to get metrics for active tunnels and on tunnel close.
In 5G session/tunnel lifetime is often short and created frequently.
The normal API becomes too slow and inaccurate when too many tunnels are created and deleted
every second.
Improvements:
- A clean ground structure to handle multiple message type in the future.
The code path for G-PDU packets is optimized for performance, representing the typical case.
Unsupported GTP-U packets enter a slow path that decodes the nature of the error.
This presents a easy hook to handle other message types in the future.
- Improved error reporting
When using traces there is more details in the tunnel descriptions.
- Updated the API with several enums.
Fixes:
- gtpu0->length field in IPv6 was computed with IPv4 header lengths in the encapsulation code.
- vec_set_len (t->rewrite, ...) size was computed with the IPv4 header size also for IPv6 tunnels.
Issues:
- This PR does not enable full support of the 3GPP specification.
In particular it only supports a single QoS/QFI flow for each tunnel.
It ignores all incoming extension header flags.
- API functions might change again when/if more support of the 3GPP TS 38.415 spec is added.
Note that I have bumped the API version to 2.1.0 as it seems to be the correct approach based on
my API changes.
Type: feature
Signed-off-by: Rune E. Jensen <runeerle@wgtwo.com>
Change-Id: I91cd2b31f2561f1b3fb1e46c4c34a5a3c71b4625
- notify app on failed connect
- avoid cleanup of ctx before transport cleanup to be able to handle
pending rx notifications.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I1b70ad45109d4c942afa1990dfce4fc44a50a637
In some anti-replay, some functions weren't using the boolean
telling if the window was huge or not. Hence, limiting the constant
propagation at compilation.
Type: fix
Change-Id: Ie5f2dda38339bb32113c6f7b2b82c82135fc92a8
Signed-off-by: Maxime Peim <mpeim@cisco.com>
Notify app that transport is closed when tcp moves to closed state.
Type: improvement
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I18cbe1ac16b1a48ecd06af4c1b5535e12e4b0e75
Make sure underlying transport connection is not removed on rescheduled
read event.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I0137a2e43aa84d9442279e036c25771aeefd207f
- test_http_static.py: If namespace creation fails, try to delete the namespace and create it again
- vpp_qemu_utils.py: Added "isinstance()" to "delete_namespace()" to match "create_namespace()"
Type: test
Change-Id: I88ff7a36f5d52816fee16283efba6af025496491
Signed-off-by: adrianvillin <avillin@cisco.com>
af_packet does not process data until the interface is UP. If after interface creation, but before it is UP, the host interfaces are flooded, then blocking case may occur - VPP interface will never be able to process the data.
If the EDGE_TRIGGERED flag is set, the event will not arrive, because nothing new is happening anymore (probably because the queue is already full).
Therefore, we need to use LEVEL_TRIGGERED (default value), which indicates that there is still unprocessed data (accumulated after interface creation, but before it was UP).
Type: fix
Signed-off-by: Artem Glazychev <artem.glazychev@xored.com>
Change-Id: Ied459fd194149d09f226bcb0a5907b3e327b148a
As a result of recent fixes, all currently stored flows of an interface
are deleted when the feature is being disabled for the interface. This
includes stopping the timer and freeing the flow entries for further
reuse. The problem is that meta information is not cleared in the flow
entries being deleted. For example, packet delta count will keep its
value. The next flow that gets one of these pool entries will already
have a non-zero packet count. So the counting of packets will start from
a non-zero value. And incorrect packet delta count will be exported for
that flow.
With this fix, clear meta information too when clearing interface state.
Also, update the corresponding test to cover this case.
Type: fix
Change-Id: I9a73b3958adfd1676e66b0ed50f1478920671cca
Signed-off-by: Alexander Chernavin <achernavin@netgate.com>
The array bounds and string overread check on GCC 12 report a dozen of
false positives that result in VPP build failures on ubuntu 22.04.
Work around this build issue by unconditionally disabling these two
warnings if C compiler is GCC 12 or newer version.
Type: fix
Signed-off-by: Jieqiang Wang <jieqiang.wang@arm.com>
Change-Id: I999e847bb625ebdf3ef5f11b11598c553f306670
GCC 12 complains about such errors while the code itself looks good.
Type: fix
Signed-off-by: Jieqiang Wang <jieqiang.wang@arm.com>
Change-Id: I021719fdbf7d9bd93a12eac76aeac8cbca13a810
DPDK added new Rx checksum flags[1] to handle cases like the virtual
drivers. Current check of flags is not strict enough for flags like
RTE_MBUF_F_RX_IP_CKSUM_NONE and will always be true no matter the
checksum in packet is good or bad.
Fix this issue by comparing the result of AND operation with the
correspinding Rx checksum flags.
Before this patch, packet trace prints the offload flags as below:
Packet Offload Flags
PKT_RX_IP_CKSUM_GOOD (0x0080) IP cksum of RX pkt. is valid
PKT_RX_IP_CKSUM_NONE (0x0090) no IP cksum of RX pkt.
PKT_RX_L4_CKSUM_GOOD (0x0100) L4 cksum of RX pkt. is valid
PKT_RX_L4_CKSUM_NONE (0x0108) no L4 cksum of RX pkt.
After this patch, packet offload flags would be like:
Packet Offload Flags
PKT_RX_IP_CKSUM_GOOD (0x0080) IP cksum of RX pkt. is valid
PKT_RX_L4_CKSUM_GOOD (0x0100) L4 cksum of RX pkt. is valid
Type: fix
[1] 5842289a54
Signed-off-by: Jieqiang Wang <jieqiang.wang@arm.com>
Change-Id: I3182022d9ccd46b2fc55bb3edfbfac9062ed7c89
Allow settings default values explicitly in positive_int_or_default and
positive_float_or_default.
It allows setting setting default 0 test retries explicitly despite it
being not positive.
Type: improvement
Signed-off-by: Dmitry Valter <d-valter@yandex-team.com>
Change-Id: Id23a9fdae0ef174eea8992c1f9fc2530aade6194
- Make framework.py classes a subset of asfframework.py classes
- Remove all packet related code from asfframework.py
- Add test class and test case set up debug output to log
- Repatriate packet tests from asf to test directory
- Remove non-packet related code from framework.py and
inherit them from asfframework.py classes
- Clean up unused import variables
- Re-enable BFD tests on Ubuntu 22.04 and fix
intermittent test failures in echo_looped_back
testcases (where # control packets verified but
not guaranteed to be received during test)
- Re-enable Wireguard tests on Ubuntu 22.04 and fix
intermittent test failures in handshake ratelimiting
testcases and event testcase
- Run Wiregard testcase suites solo
- Improve debug output in log.txt
- Increase VCL/LDP post sleep timeout to allow iperf server
to finish cleanly.
- Fix pcap history files to be sorted by suite and testcase
and ensure order/timestamp is correct based on creation
in the testcase.
- Decode pcap files for each suite and testcase for all
errors or if configured via comandline option / env var
- Improve vpp corefile detection to allow complete corefile
generation
- Disable vm vpp interfaces testcases on debian11
- Clean up failed unittest dir when retrying failed testcases
and unify testname directory and failed linknames into
framwork functions
Type: test
Change-Id: I0764f79ea5bb639d278bf635ed2408d4d5220e1e
Signed-off-by: Dave Wallace <dwallacelf@gmail.com>
Do not add ip header as that's added by tcp output and fix checksum.
Type: fix
Signed-off-by: Florin Coras <fcoras@cisco.com>
Change-Id: I9439acf5c66184af0350b1d4d7406b3feb2e79a1
