unformat_ip6_mask wasn't accounting for customized field names
when deciding if it managed to parse at least one field.
Type: fix
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: I26cab4c6828b510e277079628af5115ac43af3ff
(cherry picked from commit 126c885441)
Based on the comments in the struct, udp_encap_t_ is meant to span 2
cachelines. Due to the 64 bit alignment of dpo_id_t, the struct spanned
3 cachelines. This caused fetching ue_ip_proto to trigger an additional
cache miss. This patch rearranges the ordering of the struct fields
so that udp_encap_t_ only spans 2 cachelines as intended.
before:
(gdb) print (int)&((struct udp_encap_t_*)0)->cacheline1
$8 = 128
after:
(gdb) print (int)&((struct udp_encap_t_*)0)->cacheline1
$1 = 64
Type: fix
Signed-off-by: Vadym Martsynovskyy <vmartsyn@fb.com>
Change-Id: I066c08654d4a8ef3e2d3954e957d4c5d382b209f
(cherry picked from commit 42386fc974)
Prevent overflow if input network prefix is too small and crash on
packet #1 due to vector not being allocated/initialized.
Type: fix
Signed-off-by: Klement Sekera <ksekera@cisco.com>
Change-Id: I3494cc62ce889df48cc59cc9340b5dd70338c3a8
(cherry picked from commit f3d7bd9d4d)
If the id is invalid we cannot check whether we must free the message or
not, free it anyway.
Type: fix
Change-Id: Ie4426f601390d1e5e14c739f670e8c1e6e3aaf1e
Signed-off-by: Benoît Ganne <bganne@cisco.com>
(cherry picked from commit ff13e46215)
When recycling a graph node vnet_register_interface, it is missing an
explicit call to vlib_worker_thread_node_runtime_update(). However,
there is an implicit call to vlib_worker_thread_node_runtime_update()
via vnet_sw_interface_set_flags_helper() if it enables a new feature on
the interface for the first time. But that implicit call is not
guaranteed. For example, if an interface is created, deleted, and
created, then it may skip the implicit call to
vlib_worker_thread_node_runtime_update(). When that happens, the graph
nodes on thread 0 are not sync'ed to the worker threads. So the worker
thread's graph nodes are out of sync momentarily with the main thread's
graph nodes until some other event happens which calls for a sync is
needed. During this window, the worker thread's graph node is
vulnerable and may experience a crash.
When deleting a graph node, we never trigger a sync to the worker
thread. A patch was committed 3 years ago via
https://gerrit.fd.io/r/c/vpp/+/7523 to fix a show run crash. In
hindsight, the approach taken by 7523 is not orthogonal. While at it,
let's fix it right for both issues with a call to
vlib_worker_thread_node_runtime_update() in the appropriate place and
remove 7523.
Type: fix
Ticket: VPPSUPP-86
Fixes: gerrit 7523 / 19e9d954bd
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: Ic9472bd2d3a212dbfeceb526506ed0400983a142
(cherry picked from commit 1eae8ecb7a)
Add dpo_pool_barrier_sync/release, use them to clean up
thread-unsafe pool expansion cases.
Type: fix
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: I09299124a25f8d541e3bb4b75375568990e9b911
(cherry picked from commit 26d890eb4b)
adj_alloc (...) is not thread safe when the adj pool or combined
counter vectors expand.
Type: fix
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: I55710de6ecc083b7434e11798659cca9250c9131
(cherry picked from commit c2d2228e92)
load_balance_alloc_i(...) is not thread safe when the
load_balance_pool or combined counter vectors expand.
Type: fix
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: I7f295ed77350d1df0434d5ff461eedafe79131de
(cherry picked from commit 8341f76fd1)
Use %U and unformat_udp_port instead of %u for unformat() call for
u16 collector_port number in set_ipfix_exporter_command_fn() to
avoid corruption of other variables which can happen if unformat()
with %u is used with a 16-bit variable. This avoids crash due to
corrupted fib_index value.
Type: fix
Signed-off-by: Elias Rudberg <elias.rudberg@bahnhof.net>
Change-Id: Id54273fcc458a7f9c5aa4025aa91711f160c1c1a
(cherry picked from commit 2dca180db9)
Change in snat_ipfix_header_create() to use thread-specific
vlib_main_t *vm pointer to avoid problems with different threads
accessing the same vlib_main_t data structure. This avoids
assertion failure when vlib_time_now() is called with a vm
corresponding to a different thread.
Type: fix
Signed-off-by: Elias Rudberg <elias.rudberg@bahnhof.net>
Change-Id: I2096c1debb5688d3b97e5ed9a0ea78d94053d8b7
(cherry picked from commit 5556813fb6)
Type: fix
when the interpose is on an adj-fib and the cover is removed the adj
source will not install. this lead to no path list being found for the
interpose source and a crash. pick a drop path list in this case.
Signed-off-by: Neale Ranns <nranns@cisco.com>
Change-Id: Ied217da043926c913657080f5ffb151201225d23
(cherry picked from commit 1bf6df4ff9)
Cleanup L2/L3 mode switch to not redirect to/from ethernet-input node
as it is no longer necessary.
L2 patch should use sw_if_index for device feature enable/disable.
Type: fix
Signed-off-by: John Lo <loj@cisco.com>
Change-Id: I0f24161d027b07c188fd1e05276146f94c075710
(cherry picked from commit f415a3b53a)
When vcl_epoll_wait_handle_mq handles rx events exceeding maxevents, VPP will not signal because cursize > 0, and the remaining rx events cannot be triggered because the eventfd event has been read. Therefore, we should dequeue all events until cursize = 0. And then handle msg up to maxevents with vcl_epoll_wait_handle_mq_event and those beyond with vcl_handle_mq_event.
Type: fix
Signed-off-by: hanlin <hanlin_wang@163.com>
Change-Id: I8a0c87cb41c837deb8284b40f668cc3c7d9d6e56
Signed-off-by: hanlin <hanlin_wang@163.com>
(cherry picked from commit d0e646f689)
If a cli command is run while there are no cli session, then
cm->cli_file_pool will not be initialized and we should not try to
operate on it.
Type: fix
Change-Id: Iaea15a23f7efd5b17fab13e6c1cbb3a9a34080e0
Signed-off-by: Benoît Ganne <bganne@cisco.com>
(cherry picked from commit a58be82dda)
cxgbe PMD initializes its control channel as part of dev_configure(),
and trying to get link status prior to it will lead to a crash.
DPDK documentation loosely hints that we should not call any device
function before dev_start(), call link_state() only for the relevant
PMDs.
From DPDK API documentation:
The functions exported by the application Ethernet API to setup a device
designated by its port identifier must be invoked in the following
order:
rte_eth_dev_configure()
rte_eth_tx_queue_setup()
rte_eth_rx_queue_setup()
rte_eth_dev_start()
Then, the network application can invoke, in any order, the functions
exported by the Ethernet API to get the MAC address of a given device,
to get the speed and the status of a device physical link, to
receive/transmit [burst of] packets, and so on.
Type: fix
Change-Id: I12d2ab4d84e6bd72a9f695447e86f3222929c804
Signed-off-by: Benoît Ganne <bganne@cisco.com>
(cherry picked from commit 31eb471d0c)
Calling ethernet_set_flags() to switch interface to/from promiscuous
mode must use use hw_if_index instead of sw_if_index.
Type: fix
Signed-off-by: John Lo <loj@cisco.com>
Change-Id: I72da286b913893227e32193ee11fbbc56e04804d
(cherry picked from commit 5b960c60f6)
Fix a couple endian conversions for displaying Marker Protocol packet
in the trace
Type: fix
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: I746a67fb6143b5ad52bc4af9604ff8760dbdec9b
(cherry picked from commit 9a244b0a29)
While https://gerrit.fd.io/r/c/vpp/+/26948 fixed avoid using -1 to
index into h->free_lists[b][l] by changing the loop counter, the
check for the value of the loop counter (l < 0) cannot be trusted
to decide whether we've found a large enough object within the bin
or not. When the loop is terminated, the value of the variable l
could be ambiguous if it equals to 0 and it is never less than 0,
ie, when we bail out of the loop, we don't know if it was due to the
breaking out of the condition in
if ((s = f_size - size) >= 0)
break;
or
while (l > 0);
The fix is to explicitly set a variable when we have found a large enough
object inside the loop to be used to test whether the loop was prematurely
terminated (found == 1) or the loop just ran exhausted (found == 0)
Type: fix
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: I0161813fbd44dcba8982a767eac2e0930e9d77e3
(cherry picked from commit a5436ae251)
In search_free_list(), we have this do while loop.
do
{
l--;
f_index = h->free_lists[b][l];
f = elt_at (h, f_index);
f_size = heap_elt_size (v, f);
if ((s = f_size - size) >= 0)
break;
}
while (l >= 0);
When (l == 0), we still go back up to execute l--. Then l become -1. The
next statement is we index h->free_lists[b][-1]. After that, elt_at() would
probably cause a crash in the ASSERT.
Type: fix
Ticket: VPPSUPP-63
Signed-off-by: Steven Luong <sluong@cisco.com>
Change-Id: I617d122aa221cfdfe38f8be50f4e0f0e76e11bb5
(cherry picked from commit ec7012e51e)
When using vppcom_session* apis to setup TCP sessions in applications build outside of the VPP repository, it is necessary to set the worker_index explicitly when these apis are called from the none-VCL worker threads. An example is when data is to be sent to the TCP session that is originated from a different thread, like the main program thread or from the bin api thread. This change allows the application to set it.
Type: fix
Signed-off-by: IJsbrand Wijnands <ice@cisco.com>
Change-Id: I37f3654a49ea9a8cf3a0d3d0e672583018c12299
(cherry picked from commit 6017ff0dd7)
The path to VPP source might contain a '+' when building it
with Yocto/OpenEmbedded.
Type: fix
Signed-off-by: Ruslan Babayev <ruslan@babayev.com>
Change-Id: I205ac0de7d8726724af0e30f5b199391e05dc615
(cherry picked from commit 7f286f720d)
In hopes of restoring his +2 button...
Type: fix
Signed-off-by: Dave Barach <dave@barachs.net>
Change-Id: I2600daab5afa4334713695d1074706fbb287832f
(cherry picked from commit e891ac2f19)
vlib_free_simple_counter()
vlib_free_combined_counter()
Frees the name and two dimensional vector from the stats segment.
Type: fix
Signed-off-by: Ole Troan <ot@cisco.com>
Change-Id: If1becf7d09520ba41a3d59e2df94958ecfcf6948
(cherry picked from commit a568a19b29)
