MACVLAN devices are typically used for applications such as VRR/VRRP that
require a second MAC address (virtual). These devices have a corresponding
SVI/VLAN device -
root@TORC11:~# ip addr show vlan1002
39: vlan1002@bridge: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9152 qdisc noqueue master vrf1 state UP group default
link/ether 00:02:00:00:00:2e brd ff:ff:ff:ff:ff:ff
inet6 2001:aa:1::2/64 scope global
valid_lft forever preferred_lft forever
root@TORC11:~# ip addr show vlan1002-v0
40: vlan1002-v0@vlan1002: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9152 qdisc noqueue master vrf1 state UP group default
link/ether 00:00:5e:00:01:01 brd ff:ff:ff:ff:ff:ff
inet6 2001:aa:1::a/64 metric 1024 scope global
valid_lft forever preferred_lft forever
root@TORC11:~#
The macvlan device is used primarily for RX (VR-IP/VR-MAC). And TX is via
the SVI. To acheive that functionality the macvlan network's metric
is set to a higher value.
Zebra currently ignores the devaddr metric sent by the kernel and hardcodes
it to 0. This commit eliminates that hardcoding. If the devaddr metric
is available (METRIC_MAX) it is used for setting up the connected route
otherwise we fallback to the dev/interface metric.
Setting the macvlan metric to a higher value ensures that zebra will always
select the connected route on the SVI (and subsequently use it for next hop
resolution etc.) -
root@TORC11:~# vtysh -c "show ip route vrf vrf1 2001:aa:1::/64"
Routing entry for 2001:aa:1::/64
Known via "connected", distance 0, metric 1024, vrf vrf1
Last update 11:30:56 ago
* directly connected, vlan1002-v0
Routing entry for 2001:aa:1::/64
Known via "connected", distance 0, metric 0, vrf vrf1, best
Last update 11:30:56 ago
* directly connected, vlan1002
root@TORC11:~#
Ticket: CM-23511
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
When a local neigh is added with a MAC that is remote or absent the
neigh is kept in zebra as local/in-active. But not propagated to bgpd.
Similarly when an inactive neigh is deleted the del-msg is not propagated
to bgpd.
Without this change bgp and zebra would fall out of sync as that
bgp would not know to rerun bestpath and for it to reinstall a
known remote path for the mac-ip in question. To fix this we
now propagate inactive neigh deletes to bgpd.
Ticket: CM-23018
Testing Done:
1. evpn-min
2. manually triggered the out-of-sync state and verified the fix
Signed-off-by: Anuradha Karuppiah <anuradhak@cumulusnetworks.com>
FRR log targets are independent, so "log syslog" must not disable
"log file" output.
Fixes: #3551
Fixes: 0204baa876
Signed-off-by: David Lamparter <equinox@diac24.net>
Starting with libyang 0.16.74, we can load internally embedded yang
extensions instead of going through the file system/dlopen. Detect
support for this at build time and use if available.
NB: the fallback mechanism will go away in a short while.
Signed-off-by: David Lamparter <equinox@diac24.net>
CC lib/frr_pthread.lo
lib/frr_pthread.c:128:40: error: too many arguments to function call, expected 1, have 3
ret = pthread_setname_np(fpt->thread, fpt->os_name, NULL);
~~~~~~~~~~~~~~~~~~ ^~~~~~~~~~~~~~~~~~
/Library/Developer/CommandLineTools/SDKs/MacOSX10.14.sdk/usr/include/pthread.h:512:1: note: 'pthread_setname_np' declared here
__API_AVAILABLE(macos(10.6), ios(3.2))
Mac OS does have pthread_setname_np, but we can't use it here since it
only accepts a single argument, the thread name, and thus only works for
the current thread.
Signed-off-by: Ruben Kerkhof <ruben@rubenkerkhof.com>
Now that all daemons receive the VRF backend from zebra, we can get
rid of vrf_is_mapped_on_netns() in favor of using the more convenient
vrf_is_backend_netns() function, which doesn't require any argument.
This commit also fixes the following problem:
debian(config)# ip route 50.0.0.0/8 blackhole vrf FAKE table 2
% table param only available when running on netns-based vrfs
Even when zebra was started with the --vrfwnetns, the error
above would be displayed since the VRF FAKE didn't exist, which
would make vrf_is_mapped_on_netns() return 0 incorrectly. Using
vrf_is_backend_netns() this problem doesn't happen anymore.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Add a new field in the ZEBRA_CAPABILITIES zapi message specifying
the VRF backend in use.
For simplicity, make the zclient code call vrf_configure_backend()
to apply the received value automatically instead of requiring
the daemons to do that themselves in their zebra_capabilities()
callbacks.
Additionally, call zebra_vrf_update_all() only after sending the
capabilities message to the client, so that it will know which VRF
backend is in use when processing the VRF messages.
This commit fixes a couple of bugs in the "interface" CLI command and
associated northbound callbacks, which behave differently depending
on the VRF backend in use. Before this commit, the vrf_backend
variable would always be set to VRF_BACKEND_NETNS in the client
daemons, even when zebra was started without the --vrfwnetns option.
This could lead to inconsistent behavior and subtle bugs under
specific circumstances.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
We can make use of the vty->config variable to know when the CLI
user is in the configuration mode or not. This is much simpler
than obtaining this information from the vty node, and also a more
robust solution (the three switch statements below, for example,
were out of sync).
Also, fix a bug where vty->config wasn't being unset in the
vty_config_exit() function (bug introduced by commit f344c66ea3).
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
The CLI code uses the vty->xpath[] array and the vty->xpath_index
variables to keep track of where the user is in the configuration
hierarchy. As such, we were resetting vty->xpath_index to zero
whenever the user exited from the configuration mode in order to
keep the index valid. We weren't doing this in the vty_stop_input()
function however, which is called when the user types ^C in the
terminal. This was leading to bugs like this:
zebra> en
zebra# conf t
zebra(config)# interface eth0
zebra(config-if)# ^C
zebra# conf t
zebra(config)# interface eth0
% Configuration failed.
Schema node not found.
YANG path: /frr-interface:lib/interface[name='eth0'][vrf='default']/frr-interface:lib
To fix this, do something more clever: instead of resetting the
XPath index whenever the user exits from the configuration mode,
do that when the user enters in the configuration mode. This way
the XPath index needs to be reset in a single place only, not to
mention it's a more robust solution.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
When lyd_validate() is used with the LYD_OPT_DATA option, full YANG
validation is performed. As a side-effect to this, default nodes are
created, which is not desirable when displaying operational data
since configuration nodes can also be created. Use LYD_OPT_GET
option to resolve this problem.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
One use case for the new yang_data_list_find() function is to find
input parameters in RPC northbound callbacks easily, without the
need to iterate over the input parameters manually.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Parameters should be const whenever possible to improve code
readability and remove the need to cast away the constness of
const arguments.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
In these two functions, we were using VRF_DEFAULT instead of the
VRF ID passed as a parameter when checking if the given client
subscribed to receive default routes or not. This prevented the
"default-originate" command from ospfd/isisd from working correctly
under specific circumstances.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Since commit 3a11599c, the FRR YANG modules are embedded inside the
binaries and no longer need to be loaded from the file system. This
way, it's impossible for the FRR binaries and YANG modules to be out
of sync anymore. As such, update the suggestions of the northbound
error codes.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Instead of aborting when an incomplete xpath is given to the
nb_oper_data_iterate() function, just return an error so that the
callers have a chance to treat this error. Aborting based on invalid
user input is never the right thing to do.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
When FRR is built without the --enable-config-rollbacks option,
the nb_db_transaction_save() function does nothing and the
"transaction_id" output parameter is left uninitialized. For
this reason, all northbound clients should initialize the
"transaction_id" argument before calling nb_candidate_commit() or
nb_candidate_commit_apply() (except when a NULL pointer is given,
which is the case of the confd and sysrepo plugins).
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
We are already handling all possible four cases from the "nb_event"
enumeration, so this problem can't happen in practice. Initialize the
"ref" variable to zero to silence the warning.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Some daemons like ospfd and isisd have the ability to advertise a
default route to their peers only if one exists in the RIB. This
is what the "default-information originate" commands do when used
without the "always" parameter.
For that to work, these daemons use the ZEBRA_REDISTRIBUTE_DEFAULT_ADD
message to request default route information to zebra. The problem
is that this message didn't have an AFI parameter, so a default route
from any address-family would satisfy the requests from both daemons
(e.g. ::/0 would trigger ospfd to advertise a default route to its
peers, and 0.0.0.0/0 would trigger isisd to advertise a default route
to its IPv6 peers).
Fix this by adding an AFI parameter to the
ZEBRA_REDISTRIBUTE_DEFAULT_{ADD,DELETE} messages and making the
corresponding code changes.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
It would be nice to have the ability to access the prefix data structure
address as a block of 4 uint32_t's. This will allow me to easily/quickly
update the v6 address by 1. This will be used in subsuquent commits.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The current invocation of frr_pthread_set_name was causing it reset the os_name.
There is no need for this, we now always create the pthread appropriately
to have both name and os_name. So convert this function to a simple
call through of the pthread call now.
Before(any of these changes):
sharpd@robot ~/frr1> ps -L -p 16895
PID LWP TTY TIME CMD
16895 16895 ? 00:01:39 bgpd
16895 16896 ? 00:00:54
16895 16897 ? 00:00:07 bgpd_ka
After:
sharpd@donna ~/frr1> ps -L -p 1752
PID LWP TTY TIME CMD
1752 1752 ? 00:00:00 bgpd
1752 1753 ? 00:00:00 bgpd_io
1752 1754 ? 00:00:00 bgpd_ka
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When we start a thread we always call fpt_run and since
the last commit we know os_name is filled with something,
therefore we can just set the name on startup.
This creates this output now for zebra:
sharpd@donna ~/frr2> ps -L -p 25643
PID LWP TTY TIME CMD
25643 25643 ? 00:00:00 zebra
25643 25644 ? 00:00:00 Zebra dplane
25643 25684 ? 00:00:00 zebra_apic
sharpd@donna ~/frr2>
I removed the abstraction to frr_pthread_set_name because
it was snprintf'ing into the same buffer which was the
real bug here( the first character of os_name became null).
In the next commit I'll remove that api because
it is unneeded and was a horrible hack to get
this to work for the one place it was wanted.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
On call of frr_pthread_new, save the os_name if given,
if not given use the name passed in( shortening to fit
in available space ) and finally if the name was not
passed in use the default value.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When using getrusage, we have multiple choices about what
to call for data gathering about this particular thread of execution.
RUSAGE_SELF -> This means gather all cpu run time for all pthreads associated
with this process.
RUSAGE_THREAD -> This means gather all cpu run time for this particular
pthread.
Clearly with data gathering for slow thread as well as `show thread cpu`
it would be preferable to gather only data about the current running
pthread. This probably was the original behavior of using RUSAGE_SELF
when we didn't have multiple pthreads. So it didn't matter so much.
Prior to this change, 10 iterations of 1 million routes install/remove
from zebra would give us this cpu time for the dataplane pthread:
Showing statistics for pthread Zebra dplane thread
--------------------------------------------------
CPU (user+system): Real (wall-clock):
Active Runtime(ms) Invoked Avg uSec Max uSecs Avg uSec Max uSecs Type Thread
0 280902.149 326541 860 2609982 550 2468910 E dplane_thread_loop
After this change we are seeing this:
Showing statistics for pthread Zebra dplane thread
--------------------------------------------------
CPU (user+system): Real (wall-clock):
Active Runtime(ms) Invoked Avg uSec Max uSecs Avg uSec Max uSecs Type Thread
0 58045.560 334944 173 277226 539 2502268 E dplane_thread_loop
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
This is the start of a series of commits that will allow FRR to
be integrated into mlag.
Zebra and Pim will both need mlag state for the router. As such we will
need to provide a abstract about this state through the zapi.
This is the start of the common header that both Pim and Zebra will
be using.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
FreeBSD's libc segfaults when vsnprintf() is called with a null
format string. Add a null check before calling vsnprintf() to
resolve this problem.
Fixes#3537
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Unlike the other interface zapi messages, ZEBRA_INTERFACE_VRF_UPDATE
identifies interfaces using ifindexes and not interface names. This
is a problem because zebra always sends ZEBRA_INTERFACE_DOWN
and ZEBRA_INTERFACE_DELETE messages before sending
ZEBRA_INTERFACE_VRF_UPDATE, and the ZEBRA_INTERFACE_DELETE callback
from all daemons set the interface index to IFINDEX_INTERNAL. Hence,
when decoding a ZEBRA_INTERFACE_VRF_UPDATE message, the interface
lookup would always fail since the corresponding interface lost
its ifindex. Example (ospfd):
OSPF: Zebra: Interface[rt1-eth2] state change to down.
OSPF: Zebra: interface delete rt1-eth2 vrf default[0] index 8 flags 11143 metric 0 mtu 1500
OSPF: [EC 100663301] INTERFACE_VRF_UPDATE: Cannot find IF 8 in VRF 0
To fix this problem, use interface names instead of ifindexes to
indentify interfaces like the other interface zapi messages do.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
a distribute_ctx context pointer is returned after initialisation to the
calling daemon. this context pointer will be further used to do
discussion with distribute service. Today, there is no specific problem
with old api, since the pointer is the same in all the memory process.
but the pointer will be different if we have multiple instances. Right
now, this is not the case, but if that happens, that work will be used
for that.
distribute-list initialisation is split in two. the vty initialisation
is done at global level, while the context initialisation is done for
each routing daemon instance.
babel daemon is being equipped with a routing returning the main babel
instance.
also, a delete routine is available when the daemon routing instance is
suppressed.
a list of contexts is used inside distribute_list. This will permit
distribute_list utility to handle in the same daemon to handle more than
one context. This will be very useful in the vrf context.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
in order to enforce the vrf_id to return, from a vrf name, a check is
done on the vrf_name_to_id callback.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
This commit is the last missing piece to complete BGP LU support in bgpd. To this moment, bgpd (and zebra) supported auto label assignment only for prefixes leaked from VRFs to vpn and for MPLS SR prefixes. This adds auto label assignment to other routes types in bgpd. The following enhancements have been made:
* bgp_route.c:bgp_process_main_one() now sets implicit-null local_label to all local, aggregate and redistributed routes.
* bgp_route.c:bgp_process_main_one() now will request a label from the label pool for any prefix that loses the label for some reason (for example, when the static label assignment config is removed)
* bgp_label.c:bgp_reg_dereg_for_label() now requests labels from label pool for routes which have no associated label index
* zebra_mpls.c:zebra_mpls_fec_register() now expects both label and label_index from the calling function, one of which must be set to MPLS_INVALID_LABEL or MPLS_INVALID_LABEL_INDEX, based on this it will decide how to register the provided FEC.
Signed-off-by: Anton Degtyarev <anton@cumulusnetworks.com>
The same issue with derived enum types that was already fixed
for yang_data_new_enum was still present here, so I simply
applied the same fix.
Signed-off-by: Emanuele Di Pascale <emanuele@voltanet.io>
Just copying th const char* of the xpath means that if we
are enqueing multiple changes from a buffer, the last xpath
addedd will overwrite all of the previous references.
Copying the xpath to a buffer simplifies the API when
retrofitting the commands.
Signed-off-by: Emanuele Di Pascale <emanuele@voltanet.io>
As suggested by Renato, add error codes that are specific
to the various phases of a northbound callback. These can
be used by the daemons when logging an error. The reasoning
is that validation errors typically mean that there is an
inconsistency in the configuration, a prepare error means
that we are running out of resources, and abort/apply errors
are bugs that need to be reported to the devs.
Signed-off-by: Emanuele Di Pascale <emanuele@voltanet.io>
I accidentally put MIT headers on these; the intent was ISC. It doesn't
really make a difference, but let's get it consistent.
Signed-off-by: David Lamparter <equinox@diac24.net>
Support an optional timeout/delay for use when a workqueue
determines that it is blocked, instead of retrying immediately.
Also, schedule as an 'event' instead of a 'timer' when using
a zero timeout value.
Signed-off-by: Mark Stapp <mjs@voltanet.io>
Reorder the numbering of the Zebra message flags and document
what each flag is supposed to do.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
We never used this information and it was merely stored.
Additionally this is not something that is a flag, it's
a status.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Newer versions of clang are failing on xpath length
not being sufficiently sized to hold all possible data
that could be thrown at it.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
YANG allows lists without keys for operational data, in which case
the list elements are uniquely identified using a positional index
(starting from one).
This commit does the following:
* Remove the need to implement the 'get_keys' and 'lookup_entry'
callbacks for keyless lists.
* Extend nb_oper_data_iter_list() so that it special-cases keyless
lists appropriately. Since both the CLI and the sysrepo plugin
use nb_oper_data_iterate() to fetch operational data, both these
northbound clients automatically gain the ability to understand
keyless lists without additional changes.
* Extend the confd plugin to special-case keyless lists as well. This
was a bit painful to implement given ConfD's clumsy API, but
keyless lists should work ok now.
* Update the "test_oper_data" unit test to test keyless YANG lists in
addition to regular lists.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
ConfD and Sysrepo implement configuration transactions using a
two-phase commit protocol (prepare + abort/apply). For network-wide
transactions to work, ConfD and Sysrepo move to the second phase of
the commit protocol only after receiving the results of the first
phase from all devices involved in the transaction. If all devices
succeed in the 'prepare' phase, then all of them move to the 'apply'
phase and the transaction is committed. On the other hand, if any
device fails in the 'prepare' phase, all of them move to 'abort'
phase and the transaction is aborted.
The confd and sysrepo plugins were implementing the full
two-phase commit protocol upon receiving a request to validate
the configuration changes and allocate all resources required to
apply them (first phase). The notifications to abort or apply the
changes (second phase) were being ignored since everything was being
done in the first phase for simplicity. This wasn't a problem for
single-device transactions, but it is for transactions involving
multiple devices. Rework the code a bit to do things properly and
fix this problem.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
The "show" command will be available in the configuration mode and
all configuration subnodes. It's used to display the section of
the candidate configuration being edited, instead of displaying
the entire candidate configuration like when "show configuration
candidate" is used. The goal is to add more convenience when editing
huge configurations.
When the transactional CLI mode is not used, the candidate
configuration and the running configuration are identical, hence in
this case we can say that the "show" command displays the section
of the running configuration being edited.
Example:
ripd(config)# show
Configuration:
!
frr version 6.1-dev
frr defaults traditional
!
interface eth0
ip rip split-horizon poisoned-reverse
ip rip authentication mode md5
ip rip authentication string supersecret
!
interface eth1
ip rip receive version 1
ip rip send version 1
!
router rip
allow-ecmp
route 10.0.1.0/24
route 10.0.2.0/24
!
end
ripd(config)#
ripd(config)#
ripd(config)# interface eth0
ripd(config-if)# show
!
interface eth0
ip rip split-horizon poisoned-reverse
ip rip authentication mode md5
ip rip authentication string supersecret
!
ripd(config-if)# exit
ripd(config)#
ripd(config)#
ripd(config)# router rip
ripd(config-router)# show
!
router rip
allow-ecmp
route 10.0.1.0/24
route 10.0.2.0/24
!
ripd(config-router)#
The "show" command only works for daemons converted to the new
northbound model. vtysh support will be implemented at a later
time as it will require some level of coordination between vtysh
and the FRR daemons.
Fixes#3148.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Confirmed commits allow the user to request an automatic rollback to
the previous configuration if the commit operation is not confirmed
within a number of minutes. This is particularly useful when the user
is accessing the CLI through the network (e.g. using SSH) and any
configuration change might cause an unexpected loss of connectivity
between the user and the managed device (e.g. misconfiguration of a
routing protocol). By using a confirmed commit, the user can rest
assured the connectivity will be restored after the given timeout
expires, avoiding the need to access the router physically to fix
the problem.
When "commit confirmed TIMEOUT" is used, a new "commit" command is
expected to confirm the previous commit before the given timeout
expires. If "commit confirmed TIMEOUT" is used while there's already
a confirmed-commit in progress, the confirmed-commit timeout is
reset to the new value.
In the current implementation, if other users perform commits while
there's a confirmed-commit in progress, all commits are rolled back
when the confirmed-commit timeout expires. It's recommended to use
the "configure exclusive" configuration mode to prevent unexpected
outcomes when using confirmed commits.
When an user exits from the configuration mode while there's a
confirmed-commit in progress, the commit is automatically rolled
back and the user is notified about it. In the future we might
want to prompt the user if he or she really wants to exit from the
configuration mode when there's a pending confirmed commit.
Needless to say, confirmed commit only work for configuration
commands converted to the new northbound model. vtysh support will
be implemented at a later time.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Introduce frr-ripngd.yang, which defines a model for managing the
FRR ripngd daemon.
Update the 'frr_yang_module_info' array of ripngd with the new
'frr-ripngd' module.
Add two new files (ripng_cli.[ch]) which should contain all ripngd
commands converted to the new northbound model. Centralizing all
commands in a single place will facilitate the process of moving
the CLI to a separate program in the future.
Add automatically generated stub callbacks in
ripng_northbound.c. These callbacks will be implemented gradually
in the following commits.
Add the confd.frr-ripngd.yang YANG module with annotations specific
to the ConfD daemon.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This reverts commit 48944eb65e.
We're using GNU C, not ISO C - and this commit triggers new (real)
warnings about {0} instead of bogus ones about {}.
Signed-off-by: David Lamparter <equinox@diac24.net>
We sometimes store ifindex information in the NEXTHOP_TYPE_IPV[4|6]
so let's let us display that information as well when dumping
a nexthop.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
There are cases where the passed parameter for a vty command is either
an interface name or an ip address. Because the interface name can be a
number, and because the user may want to use a number to define an IP (
for instance 'ping 0' is valid from shell purpose), there is a choice
that needs to be done at frr level. either from the application point of
view, the interface name will be priorized, or each number will be
considered as an ip address. In that commit, the inet_aton procedure is
replaced with the inet_pton procedure that ignores ips with just a
number.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
The northbound infrastructure for operational data was subpar compared
to the infrastructure for configuration data. This commit addresses most
of the existing problems, making it possible to write operational-data
callbacks for more complex YANG models.
Summary of the changes:
* Add support for nested YANG lists.
* Add support for leaf-lists.
* Add support for leafs of type "empty".
* Introduce the "show yang operational-data XPATH" command, and write an
unit test for it. The main purpose of this command is to make it
easier to test the operational-data northbound callbacks.
* Introduce the nb_oper_data_iterate() function, that can be used
to iterate over operational data. Make the CLI and sysrepo use this
function.
* Since ConfD has a very peculiar API, it can't reuse the
nb_oper_data_iterate() like the other northbound clients. In this
case, adapt the existing ConfD callbacks to support the new features
(and make some performance improvements in the process).
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Prevent the confd plugin from subscribing to configuration changes on a
data tree that contains only state data.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
A YANG list that contains both configuration and state data must have
the following callbacks: create(), delete(), get_next(), get_keys()
and lookup_entry().
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
* Rename yang_snodes_iterate() to yang_snodes_iterate_subtree() and
expose it in the public API.
* Rename yang_module_snodes_iterate() to yang_snodes_iterate_module().
* Rename yang_all_snodes_iterate() to yang_snodes_iterate_all().
* Make it possible to stop the iteration at any time by returning
YANG_ITER_STOP in the iteration callbacks.
* Make the iteration callbacks accept only one user argument and not
two.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
In some cases it will be necessary to load all FRR native modules.
Examples:
* vtysh needs to load all YANG modules so that it can manipulate data
from all daemons.
* The gen_northbound_callbacks tool will need to load all YANG modules
since augmentations from one module can have an effect in the required
northbound callbacks of other modules.
The new yang_module_load_all() function provides this functionality.
As a side note, the "frr_native_modules" will need to be updated every
time we add a new YANG module to FRR.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
For convenience, make yang_dnode_free() remove the entire data tree and
not only the data node given as a parameter.
Also, add a null-pointer check on nb_config_replace() before calling
yang_dnode_free().
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
By default the data nodes created by yang_dnode_new() could contain
only configuration data (LYD_OPT_CONFIG). Add a 'config_only' option
to yang_dnode_new() so that it can create data nodes containing both
configuration and state data.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Prefetching the schema node when creating yang_data structures is
expensive, and in most cases we don't need that information. In that case,
fetch the schema information only when necessary to improve performance
when fetching operational data.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
A while ago all FRR configuration commands were converted to use the
QOBJ infrastructure to keep track of configuration objects. This
means the configuration lock isn't necessary anymore because the
QOBJ code detects when someones tries to edit a configuration object
that was deleted and react accordingly (log an error and abort the
command). The possibility of accessing dangling pointers doesn't
exist anymore since vty->index was removed.
Summary of the changes:
* remove the configuration lock and the vty_config_lockless() function.
* rename vty_config_unlock() to vty_config_exit() since we need to
clean up a few things when exiting from the configuration mode.
* rename vty_config_lock() to vty_config_enter() to remove code
duplication that existed between the three different "configuration"
commands (terminal, private and exclusive).
Configuration commands converted to the new northbound model don't
need the configuration lock either since the northbound API also
detects when someone tries to edit a configuration object that
doesn't exist anymore.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
When editing the candidate configuration, the northbound must ensure
that either all changes made by a command are accepted or none are.
This is done to prevent inconsistent states where only parts of a
command are applied in the event any error happens.
The previous API for converted commands, the nb_cli_cfg_change()
function, required callers to pass an array containing all changes
that needed to be applied in the candidate configuration. The
problem with this API is that it was very inconvenient for complex
commands, which change different configuration options depending
on several factors. This required users to manipulate the array
of configuration changes using low-level primitives, making it
complicated to implement some commands.
To solve this problem, introduce a new API based on the two following
functions:
- nb_cli_enqueue_change()
- nb_cli_apply_changes()
The first function is used to enqueue configuration changes, one
at time. Then the nb_cli_apply_changes() function is used to apply
all the enqueued configuration changes.
To implement this, a static-sized array was allocated in the "vty"
structure, along with a counter of enqueued changes. This eliminates
the need to declare an array of configuration changes in every
converted CLI command, simplifying things quite considerably.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Add the "abort_if_not_found" parameter to the yang_dnode_get_entry()
function instead of always aborting when an user pointer is not
found. This will make it possible, for example, to use this function
during the validation phase of a configuration transaction. Callers
will only need to check if the function returned NULL or not,
since new configuration objects (if any) won't be created until
the NB_EV_APPLY phase of the transaction.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
In some cases it might be desirable to obtain the schema name of
a libyang data node. Introduce the yang_dnode_get_schema_name()
function for this purpose.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This fixes an infinite loop that happened every time the connection
to the confd daemon was lost. Deactivate the confd module when
that happens to fix the infinite loop. This is only a temporary
workaround, in the long term we need to add a connection retry timer
to reestablish the connection to the confd daemon once it's back.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
the netns discovery process executed when vrf backend is netns, allows
the zebra daemon to dynamically change the default vrf name value. This
option is disabled, when the zebra is forced to a default vrf value with
option -o.
PR=61513
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
This bakes our YANG models straight into the library/daemons, so they
don't need to be loaded from /usr/share/yang. This makes the
installation quite a bit more robust, as well as gets us halfway to
running uninstalled. (The other half is baking in the extension type
module.)
The /usr/share/yang directory is still searched as a fallback, as well
as for the experimental YANG model translator. This is likely to stay
as is for the time being.
Signed-off-by: David Lamparter <equinox@diac24.net>
Display following Per MAC and Neigh's output:
If duplicate address detection is under process,
display detection start time and detection count.
If duplicate address detection detected an address
as duplicate, display detection time and duplicate
status.
Signed-off-by: Chirag Shah <chirag@cumulusnetworks.com>
Duplicate address detection configuration clis
under bgp l2vpn evpn config mode.
- Enabled/Disable (global knob) for feature.
- Configure cli for duplicate detection action
freeze and freze until time (auto-recovery).
Signed-off-by: Chirag Shah <chirag@cumulusnetworks.com>
if zebra is not started, then vrf identifiers are not available. This
prevents import/exportation to be available. This commit permits having
import/export available, even when zebra is not started.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
This commit introduces lib/id_alloc, which has facilities for both an ID number
allocator, and less efficient ID holding pools. The pools are meant to be a
temporary holding area for ID numbers meant to be re-used, and are implemented
as a linked-list stack.
The allocator itself is much more efficient with memory. Based on sizeof
values on my 64 bit desktop, the allocator requires around 155 KiB per
million IDs tracked.
IDs are ultimately tracked in a bit-map split into many "pages." The
allocator tracks a list of pages that have free bits, and which sections
of each page have free IDs, so there isn't any scanning required to find
a free ID. (The library utility ffs, or "Find First Set," is generally a
single CPU instruction.) At the moment, totally empty pages will not be
freed, so the memory utilization of this allocator will remain at the
high water mark.
The initial intended use case is for BGP's TX Addpath IDs to be pulled
from an allocator that tracks which IDs are in use, rather than a free
running counter. The allocator reserves ID #0 as a sentinel value for
an invalid ID numbers, and BGP will want ID #1 reserved as well. To
support this, the allocator allows for IDs to be explicitly reserved,
though be aware this is only practical to use with low numbered IDs
because the allocator must allocate pages in order.
Signed-off-by Mitchell Skiba <mskiba@amazon.com>
ipv6 distribute-list name picked up was not the correct one. the
parameter number is modified accordingly.
Also, the unconfiguration of distribute-list ipv6 was conflicting with
other daemon, thus making impossible the unconfigration. The command has
been split to be specific to ipv6 distribute-list.
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
This duplicates itself N times since it's not wrappered in a vtysh
command. In lieu of doing that, just remove the message, it's not really
necessary.
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
It's been a year since we added the new optional parameters
to instantiation. Let's switch over to the new name.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The frr-interface YANG module models interfaces using a YANG list keyed
by the interface name and the interface VRF. Interfaces can't be keyed
only by their name since interface names might not be globally unique
when the netns VRF backend is in use. When using the VRF-Lite backend,
however, interface names *must* be globally unique. In this case, we need
to validate the uniqueness of interface names inside the appropriate
northbound callback since this constraint can't be expressed in the
YANG language. We must also ensure that only inactive interfaces can be
removed, among other things we need to validate in the northbound layer.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
Introduce frr-interface.yang, which defines a model for managing FRR
interfaces.
Update the 'frr_yang_module_info' array of all daemons that will
implement this module.
Add automatically generated stub callbacks in if.c. These callbacks will
be implemented in the following commit.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This plugin leverages the northbound API to integrate FRR with Sysrepo,
a YANG-based configuration and operational state data store.
The plugin is linked to the libsysrepo library and communicates with
the sysrepod daemon using GPB (Google Protocol Buffers) over AF_UNIX
sockets. The integration consists mostly of glue code that calls the
appropriate FRR northbound callbacks in response to events triggered
by the sysrepod daemon (e.g. request to change the configuration or to
fetch operational data).
To build the sysrepo plugin, provide the --enable-sysrepo option to the
configure script while building FRR (the libsysrepo library needs to be
installed in the system).
When installed, the sysrepo plugin will be available for all FRR daemons
and can be loaded using the -M (or --module) command line option.
Example: bgpd -M sysrepo.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
This plugin leverages the northbound API to integrate FRR with the ConfD
management agent.
The plugin is linked to the libconfd library and communicates with the
confd daemon using local TCP sockets. The integration consists mostly
of glue code that calls the appropriate FRR northbound callbacks in
response to events triggered by the confd daemon (e.g. request to change
the configuration or to fetch operational data).
By integrating FRR with the libconfd library, FRR can be managed using
all northbound interfaces provided by ConfD, including NETCONF, RESTCONF
and their Web API.
The ConfD CDB API is used to handle configuration changes and the ConfD
Data Provider API is used to provide operational data, process RPCs and
send notifications. Support for configuration management using the ConfD
Data Provider API is not available at this point.
The ConfD optional 'get_object()' and 'get_next_object()' callbacks were
implemented for optimal performance when fetching operational data.
This plugins requires ConfD 6.5 or later since it uses the new leaf-list
API introduced in ConfD 6.5.
To install the plugin, the --enable-confd option should be given to the
configure script, specifying the location where ConfD is installed.
Example: ./configure --enable-confd=/root/confd-6.6
When installed, the confd plugin will be available for all FRR daemons
and can be loaded using the -M (or --module) command line option.
Example: zebra -M confd.
Signed-off-by: Renato Westphal <renato@opensourcerouting.org>
* Cast when assigning should be to uint16_t
* Restored comment documenting strange behavior
* Further increased PREFIX_STRLEN to 80 chars
Signed-off-by: Quentin Young <qlyoung@cumulusnetworks.com>
The CMD_SUCCESS_DAEMON case should be excluded from storing the command line
that we think failed.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
We weren't cleaning up some files (a whole lot of python foobar) and had
some files in the dist tarball that don't quite belong there.
Signed-off-by: David Lamparter <equinox@diac24.net>
The ->hash_cmp and linked list ->cmp functions were sometimes
being used interchangeably and this really is not a good
thing. So let's modify the hash_cmp function pointer to return
a boolean and convert everything to use the new syntax.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
libunwind provides an alternate to backtrace() for printing out the call
stack of a particular location. It doesn't use the frame pointer, it
goes by the DWARF debug info. In most cases the traces have exactly the
same information, but there are some situations where libunwind traces
are better.
(On some platforms, the libc backtrace() also uses the DWARF debug info
[e.g.: ARM backtraces are impossible without it] but this is not the
case everywhere, especially not on BSD libexecinfo.)
Signed-off-by: David Lamparter <equinox@diac24.net>
Allow the modification of whether or not we will allow
BUM flooding on the vxlan bridge. To do this allow
the upper level protocol to specify via the ZEBRA_VXLAN_FLOOD_CONTROL
zapi message.
If flooding is disabled then BUM traffic will not be forwarded
to other VTEP's.
Signed-off-by: Vivek Venkatraman <vivek@cumulusnetworks.com>
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When reading in config files and we have failures on multiple
lines actually note the actual failure lines and return them.
This fixes an issue where we stopped counting errors after
the first one and we got missleading line numbers that
did not correspond to the actual problem.
This is fixed:
sharpd@donna ~/frr> sudo /usr/lib/frr/pimd --log=stdout -A 127.0.0.1 -f /etc/frr/pimd.conf
2018/10/11 09:41:01 PIM: VRF Created: default(0)
2018/10/11 09:41:01 PIM: pim_vrf_enable: for default
2018/10/11 09:41:01 PIM: zclient_lookup_sched_now: zclient lookup immediate connection scheduled
2018/10/11 09:41:01 PIM: zclient_lookup_new: zclient lookup socket initialized
2018/10/11 09:41:01 PIM: pimd 6.1-dev starting: vty@2611
2018/10/11 09:41:01 PIM: [EC 100663304] ERROR: No such command on config line 2: inteface lo
2018/10/11 09:41:01 PIM: [EC 100663304] ERROR: No such command on config line 3: ip igmp
2018/10/11 09:41:01 PIM: [EC 100663304] ERROR: No such command on config line 4: ip igmp join 224.1.1.1 13.13.13.2
^C2018/10/11 09:45:09 PIM: Terminating on signal SIGINT
2018/10/11 09:45:09 PIM: VRF Deletion: default(0)
Fixes: #3161
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Don't allocate threads in the stack, but use the standardized
`thread_get` and `thread_add_unused` to avoid creating corner cases in
the thread API.
This fixes a thread mutex memory leak in FreeBSD.
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
Two important changes:
* Centralize the thread teardown procedure;
* Save and restore thread mutex context to avoid losing the memory
pointer;
Signed-off-by: Rafael Zalamena <rzalamena@opensourcerouting.org>
The compiler.h header provides us with some useful macro's
that we are using in the system. We do not know exactly
where the CPP_NOTICE and CPP_WARN macros are used but
they can move around. Place this header early in the
build then.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
This cleans up watchfrr to be more "normal" like the other daemons in
terms of what it does in main(), i.e. using the full frr_*() call set.
Also, this changes the startup behaviour on watchfrr to stay attached on
the daemon's parent process until startup is really complete. This
should allow removing the "watchfrr.started" hack at some point.
Signed-off-by: David Lamparter <equinox@diac24.net>
This makes libfrr.so executable to print its version info. This is
useful if you need to check your libfrr.so matches your daemons.
Signed-off-by: David Lamparter <equinox@diac24.net>
This option can be used to get statically linked binaries.
Note: libfrr.la is removed from modules' library dependency list. This
is intentional and explained in a comment in lib/subdir.am.
Signed-off-by: David Lamparter <equinox@diac24.net>
Overview:
Coverity points a copy-paste error in the Red-Black tree implementation. The
RB tree code is based on the OpenBSD implementation, so at first glance, it
is a strong point for thinking twice before touching anything.
Details:
The code is an augmented RB tree implementation [1], which adds to RB trees
the possibility of using a callback on every node update for updating per-node
associated metainformation. The bug is clear once checking other places where
the callback is called.
Impact:
- FRR: no impact, because the "augmented" capability is not being used.
- OpenBSD [2]: it seems there is no impact, at least in the 'src' repository.
Additional observations:
- If the "augmented" capability is not used, the code could run faster (at
every operation on a node the callback is checked for not being NULL). May
be branch prediction could be enough for those extra operations being
negligible on most processors in use.
[1] http://kaba.hilvi.org/pastel-1.3.0/pastel/sys/redblacktree.htm
[2] GH mirror: https://github.com/openbsd/src/blob/master/sys/kern/subr_tree.c
Signed-off-by: F. Aragon <paco@voltanet.io>
Keep track of how often route-maps are applied and
how often each clause of a route-map is applied.
This change showed that `show route-map` was outputting
odd data so fix that output and add in the applied
times too.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Wrapper the get/set of the table->info pointer so that
people are not directly accessing this data.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When entering a interface name and you fat-finger it
actually display some useful information about the vrf
we are in.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
When the "call" CLI is executed from with-in a route-map that is already in use,
there is a need to get the route-map clients to re-evalute the clauses defined
by both the parent route-map, as well as the child route-map.
The existing callbacks, add_hook() and delete_hook() can be used by the lib to
inform the clients when the "call" is configured and unconfigured.
Signed-off-by: NaveenThanikachalam <nthanikachal@vmware.com>
Redundant parentheses surrounding declarator removed.
Can be detected via static analysis with e.g.
./configure CFLAGS=-Wredundant-parens CC=clang
Signed-off-by: F. Aragon <paco@voltanet.io>
OS-level yield is generally a bad and possibly dangerous idea. If the
thread should be suspended, there should always be something to wait on,
or it turns into busy waiting. And if it's "just giving something else
the chance to run" - that's the kernel's job to determine, and the
kernel will do so while considering priorities, cgroups, and whatnot.
Let it do its job.
Signed-off-by: David Lamparter <equinox@diac24.net>
All I can see is an unneccessary complication. If there's some purpose
here it needs to be documented...
Signed-off-by: David Lamparter <equinox@diac24.net>
Corrections so that the BGP daemon can work with the label manager properly
through a label-manager proxy. Details:
- Correction so the BGP daemon behind a proxy label manager gets the range
correctly (-I added to the BGP daemon, to set the daemon instance id)
- For the BGP case, added an asynchronous label manager connect command so
the labels get recycled in case of a BGP daemon reconnection. With this,
BGPd and LDPd would behave similarly.
Signed-off-by: F. Aragon <paco@voltanet.io>
Allow at timer wheel creation time the ability to specify a
name for what we want the 'show thread cpu' to show up as.
Modify pim to note this.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Allow the user to specify a run name for display in
'show thread cpu' that is different than the function
name we are calling.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
FreeBSD supports pthread_set_name_np() too. Also, pthread_set_name_np()
returns void. And NetBSD has pthread_setname_np() with an extra arg...
Signed-off-by: David Lamparter <equinox@diac24.net>
Need this to get CMSG_SPACE/CMSG_LEN on Solaris.
Also, AC_GNU_SOURCE is deprecated, AC_USE_SYSTEM_EXTENSIONS does that.
Signed-off-by: David Lamparter <equinox@diac24.net>
config.h (or, transitively, zebra.h) must be the first include file
listed for autoconf things like _GNU_SOURCE and _POSIX_C_SOURCE to work
correctly.
Signed-off-by: David Lamparter <equinox@diac24.net>
ASAN/MSAN/TSAN flags need to be in CFLAGS and LDFLAGS; the latter links
the correct compiler-dependent library. Also, the configure switch was
broken (--disable-... would enable the sanitizer.)
Signed-off-by: David Lamparter <equinox@diac24.net>
Since we're now building through one large Makefile, we can easily put
things with their daemons and crossreference nicely.
Signed-off-by: David Lamparter <equinox@diac24.net>
Clang was thinking the random level could be negative. (And, no, I
couldn't figure that out by reading its output... trial and error this
was.)
Signed-off-by: David Lamparter <equinox@diac24.net>
Add a TAILQ_POP_FIRST so Clang understands it's the same item that is
getting removed from the list.
Signed-off-by: David Lamparter <equinox@diac24.net>
Auto-detect if pthread_condattr_setclock is available and if
it is not allow the code to compile around the issue.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
The vty_prefix_list_install function was modifying the prefix to match the
specified prefix length and warning in the log file. Modify
code to use zlog_info as that a warn implies that something has
gone terribly wrong. Additionally display to the terminal as
well so that user can get immediate feedback from something
that they can correct.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Solution :
The following procedures would be performed :
1. Verify if the pid file for each daemon is present or not. If the file is not present, that means the
daemon is getting instantiated for the first time. So let it go ahead.
If the file is present proceed to point ‘2’.
2. Try fetching the properties of the pid file.
3. If it has RW lock, that means one instance of this the daemon is already running.
So stop moving ahead and do exit() else let it go ahead. Please note all above procedure happen at
the initial state of daemon’s instantiation, much before it starts any session with other
process/allocates resources etc.. and this verification do not have any impact of any
operations done later, if the verification succeeds.
Signed-off-by: bisdhdh sadhub@vmware.com
For OpenFabric operation, we need to be able to install routes via
interfaces without any IPv4 addresses configured. Introduce a flag
ZEBRA_FLAG_ONLINK which upper protocols can set on a route they send
towards zebra, to force the nexthops to be considered onlink.
Signed-off-by: Christian Franke <chris@opensourcerouting.org>
fabricd is built using the sources of isisd. To allow differentiation
in the code, -DFABRICD=1 is added to its preprocessor flags.
Signed-off-by: Christian Franke <chris@opensourcerouting.org>
The ZEBRA_IPV4_ROUTE_IPV6_NEXTHOP_ADD zapi message has no creators and
no handlers. Let's just remove.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Move the aggregate pointer from the route_node into agg_node
so that people using struct route_node will see a savings
in data size.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Add a abstraction for `struct route_node` and `struct route_table`
such that we can have an aggregate route_node and table. This
is because only bgp/rfapi and ripng use the aggregate data pointer
in `struct route_node`. For full route tables other routing
protocols and tables are paying a 8 byte overhead per node.
A full bgp table ends up being ~1.2 million routes in bgp
and zebra. This is not an insiginificant amount of data.
So create the data structures for this replacement, but
do not replace the aggregate pointer yet. This is because
later commits will convert rfapi and ripng over to this
new data, and finally we'll move the aggregate pointer.
Signed-off-by: Donald Sharp <sharpd@cumulusnetworks.com>
Fix CLANG warning:
Report for if.c | 2 issues
===============================================
< WARNING: else is not generally useful after a break or return
< #390: FILE: /tmp/f1-28557/if.c:390:
Signed-off-by: Thibaut Collet <thibaut.collet@6wind.com>
Problem reported that some bgp and ospf json commands did not return
any json output at all if the bgp/ospf instance did not exist.
Additionally, some bgp and ospf json commands did not return any json
output if the instance existed but no neighbors were defined. This
fix makes these commands more consistent in returning empty braces for
json output and issue a message if not using json output. Additionally,
made the flag "use_json" a bool to make it consistent since previously,
it had been defined as an int, char, u_char, and bool at various places.
Ticket: CM-21040
Signed-off-by: Don Slice <dslice@cumulusnetworks.com>
This crash occurs only with netns implementation.
vrf meaning is different regarging its implementation (netns or
vrf-lite)
- With vrf-lite implementation vrf is a property of the interface that
can be changed as the speed or the state (iproute2 command: "ip link
set dev IF_NAME master VRF_NAME"). All interfaces of the system are in
the same netns and so interface name is unique.
- With netns implementation vrf is a characteristic of the interface
that CANNOT be changed: it is the id of the netns where the interface
is located. To change the vrf of an interface (iproute2 command to
move an interface "ip netns exec VRF_NAME1 ip link set dev IF_NAME
netns VRF_NAME2") the interface is deleted from the old vrf and
created in the new vrf.
Interface name is not unique, the same name can be present in the
different netns (typically the lo interface) and search of interface
must be done by the tuple (interface name, netns id).
Current tests on the vrf implementation (vrf-lite or netns) are not
sufficient. In some cases (for example when an interface is moved from
a vrf X to the default vrf and then move back to VRF X) we can have a
corruption message and then a crash of zebra.
To avoid this corruption test on the vrf implementation, needed when an
interface changes, has been rewritten:
- For all interface changes except deletion the if_get_by_name function,
that checks if an interface exists and creates or updates it if
needed, is changed:
* The vrf-lite implementation is unchanged: search of the interface
is based only on the name and update the vrf-id if needed.
* The netns implementation search of the interface is based on the
(name, vrf-id) tuple and interface is created if not found, the
vrf-id is never updated.
- deletion of an interface (reception of a RTM_DELLINK netlink message):
* The vrf-lite implementation is unchanged: the interface
information are cleared and the interface is moved to the default
vrf if it does not belong to (to allow vrf deletion)
* The netns implementation is changed: only the interface
information are cleared and the interface stays in its vrf to
avoid conflict with interface with the same name in the default
vrf.
This implementation reverts (partially or totally):
commit 393ec5424e ("zebra: fix missing node attribute set in ifp")
commit e9e9b1150f ("lib: create interface even if name is the same")
commit 9373219c67 ("zebra: improve logs when replacing interface to an
other netns")
Fixes: b53686c52a ("zebra: delete interface that disappeared")
Signed-off-by: Thibaut Collet <thibaut.collet@6wind.com>
Signed-off-by: Philippe Guibert <philippe.guibert@6wind.com>
